TH1638

MODELING OF ROTARY CEMENT KILNS
Submitted in partial fulfillment of the requirements

for the degree of
DOCTOR OF PHILOSOPHY
by
Kaustubh S. Mujumdar
Roll no. 03402701

Supervisor: Prof. Anurag Mehra
DEPARTMENT OF CHEMICAL ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY
BOMBAY 400 076
2008
Thesis entitled: MODELING OF ROTARY CEMENT KILNS
by KAUSTUBH S. MUJUMDAR
is approved for the degree of DOCTOR OF PHILOSOPHY
Examiners
Supervisors
Chairman
Date
Place
INDIAN INSTITUTE OF TECHNOLOGY, BOMBAY, INDIA
CERTICATE OF COURSE WORK
This is to certify that Mr. Kaustubh S. Mujumdar was admitted to the candidacy of the
Ph. D. Degree on Jan 2004, after successfully completing all the courses required or the
Ph. D. Degree programme. The details of the course work done are given below.
Sr. No Course No. Course Name Credits

1 CL 601 Advanced Transport Phenomena 6.0
2 CL 649 Reaction Engineering in Dispersed 6.0
Phase System
3 CL 704 Seminar 4.0
4 CLs 801 Communication and Presentation 4.0
Skills
I.I.T Bombay Dy. Registrar (Academic)

Date:
Abstract
Rotary cement kilns are key equipment in cement industry used for converting calcineous
raw meal into cement clinkers. In this work, a combination of reaction engineering model and
a CFD model was used to build adequately complete representation of practical cement kilns.
In the first part of thesis, development of reaction engineering based models for rotary
cement kilns was reported. The model was formulated by coupling three models viz. model
for simulating variation of bed height in the kiln, model for simulating reactions and heat
transfer in the bed region and model for simulating coal combustion and heat transfer in the
freeboard region of cement kilns. Melting and formation of coating within the kiln were
accounted. Combustion of coal in the freeboard region was modeled by accounting
devolatilization, finite rate gas phase combustion and char reaction. The simulated results
were validated with the available data from 3 industrial kilns. The model was then used to
understand influence of various design and operating parameters on kiln performance.
Several ways of reducing energy consumption in kilns were then computationally
investigated. The model was also used to propose and to assess a practical solution of using a
secondary shell to reduce energy consumption in rotary cement kilns. Simulation results
indicated that varying kiln operating variables viz. solid flow rate or RPM can result only in
minor changes in kiln energetics. Use of secondary shell over kiln and energy recovery by
passing air through the annular gap between the two, appears to be a promising way for
significant energy saving. The developed model and the presented results will be useful for
enhancing performance of rotary cement kilns.
Since the rotary kiln is closely coupled with associated equipments, models for pre-heater,
calciner and cooler were also developed and coupled with kiln model to develop an integrated
simulator for clinker manufacture in cement industry. The model for pre-heater cyclones was
formulated by assuming solids and gas to be completely back mixed. The heat transfer
coefficient between particle laden gas and walls cyclones was calculated using empirical
correlation. The model for calciner was formulated based on Eulerian-Lagrangian approach.
The coal particles and raw meal particles were considered as discrete phases and gas phase
was assumed to be completely back mixed. Clinker cooler was modeled by assuming solids
and gas in plug flow. The heat transfer between solids and air in clinker cooler was predicted
i
via empirical correlations. The simulation results using the simulator have predicted results
reasonable with industrial observations.
It is important to note that because of inherent limitations of 1D framework, the developed
reaction engineering model will not able to capture influence of burner design and key
operating parameters like ratio of swirl or axial air etc. on performance of cement kilns.
Computational fluid dynamics based models were therefore developed for rotary cement
kilns on second front to achieve this. As a starting point, efforts were initiated to develop
CFD models for motion of solids and heat transfer in transverse section of rotating cylinder.
A systematic study of various physical parameters like diameter of solids, restitution
coefficient, maximum packing limit, angle of internal friction, frictional viscosity on
transverse flow revealed that frictional viscosity has the most significant impact on flow of
solids in transverse plane. However, the developed models for frictional viscosity did not
match the available experimental data. Use of simpler models like pseudo-homogeneous
assumption for the solids phase were found to give results similar to those with much more
complicated models. CFD models to predict heat transfer in partially filled rotating cylinders
indicated that assuming solids as pseudo fluids with constant viscosity in the bed region was
reasonable to predict temperature gradients in the bed region. Therefore the CFD model for
the kiln was developed assuming solids as pseudo fluids.
The development of CFD model for rotary cement kilns was then reported. Separate CFD
models were developed for the bed and the free board regions of the rotary kiln. The
individual models were coupled with each other via mass and energy communication through
the common boundaries. Appropriate source and sink terms were formulated to account for
such mass and heat transfer communication between the bed and the freeboard regions. These
models were solved iteratively using suitable under relaxation factors. The simulated results
were compared with the available data from industrial kilns and were seen to capture cement
kiln behavior reasonably well. The model was also used to examine influence of burner
characteristics on burning zone as well as overall performance of the kiln. The developed
approach, CFD model and simulation results not only are able to predict the behavior of
industrial kiln reasonably but also able to predict influence of burner design and operating
parameters on the kiln performance .
Overall, the developed model and presented results will be helpful in developing better
understanding of clinker manufacture in cement plants.
ii
Contents
Abstract i
Notations v
1. Introduction 1
1.1. Introduction 2
1.2. Rotary Cement Kilns 3
1.3. Motivation of present work 5
1.4. Methodology adopted 6
1.5. Organization of thesis 8
2. Key issues in Modeling Cement Kilns 10
2.1. Introduction 11
2.2. Flow of solids in rotary cement kilns 12
2.2.1. Flow of solids in axial direction 12
2.2.2. Flow of solids in transverse direction 19
2.3. Clinkerization reactions 22
2.4. Melt and coating formation 25
2.5. Heat transfer 26
2.6. Freeboard region of cement kilns 29
2.7. Summary of key issues in modeling cement kilns 31
2.8. Performance models for rotary cement kilns 32
2.8.1. One dimensional models 32
2.8.2. CFD models 33
3. Reaction Engineering based Models 35
3.2. Computational model and solution methodology 37
3.2.1. Model for calculating bed height in rotary kilns 37
3.2.2. Bed model 39
3.2.3. Freeboard model 52
3.2.4. Solution methodology 57
3.3. Results and Discussion 58
3.3.1. Validation of bed model 59
3.3.2. Simulation of rotary cement kilns 61
3.3.3. Influence of key design and operating parameters 68
3.3.4. Energy consumption in rotary cement kilns 70
3.3.5. Use of secondary shell 75
3.4. Conclusions 77
4. Applications of reaction engineering based models to clinker manufacture 79
4.2. Key issues and modeling approach 80
4.2.1. Cyclone pre-heater 81
4.2.2. Calciner 82
4.2.3. Rotary kiln 83
4.2.4. Clinker cooler 83
4.3. Computational model for pre-heater, calciner, cooler 84
4.3.1. Pre-heater model 84
4.3.2. Calciner model 89
4.3.3. Kiln model 99
iii
4.3.4. Clinker cooler model 99
4.4. Integrated model 104
4.5. RoCKS: Rotary Cement Kiln Simulator 109
4.6. Simulation results 112
4.7. Conclusions 126
5. Computational fluid dynamics based models for transverse section in
rotating kilns 127
5.2. Motion of solids in transverse section of rotary kilns 129
5.2.1. Model description 131
5.2.2. Geometry, grid and Solver 133
5.2.3. Numerical Experiments and results 134
5.3. Heat transfer in transverse section of rotating kiln 144
5.3.1. Model description 144
5.3.2. Numerical Experiments and results 147
6. Computational fluid dynamics based models for rotary cement kilns 154
6.2. Computational Model 155
6.2.1. CFD models for cement kilns 156
6.2.2. Bed model 157
6.2.3. Freeboard Model 160
6.2.4. Mass transfer from bed to freeboard 164
6.2.5. Methodology adopted 165
6.3. Results and discussion 167
6.3.1. Coupling of bed and freeboard region 167
6.3.2. Simulation of rotary cement kilns 171
6.3.3. Effect of burner operational parameters on the kiln performance 176
7. Conclusions 180
References 184
List of Publications
Acknowledgements
iv
NOTATIONS
Acl cross sectional area covered by the charge, m2

Acyi Internal surface area of cyclone, K
Acyo External surface area of cyclone, K
Ap Surface area of coal particle, m2
L
Ap Surface area of solid particle, m2
a Surface area per unit volume, m2/m3
ACGW Convection area gas to wall, m2
ACGB Convection area from gas to bed, m2
ACWB Conduction area from wall to bed, m2
AP Particle surface area, m2
ARGB Radiation area from gas to bed, m2
ARWB Radiation area from wall to solids, m2
A0 Devolatilisation rate constant
CC Concentration of CaCO3 particle, kg/m3
Ck Concentration of the component in the bed, kg/m3
Cp Specific heat of the bed, kJ/ kg K
CT Coating thickness, m
CT, max Maximum coating thickness, m
dc Inner diameter of cyclone, m
dp Radius of particle, m
De Equivalent Diameter, m
Ecoal Energy given by coal in the kiln, kJ/ kg clinker
Ej Activation energy of the jth reaction, kJ/ mol
E1 Energy of activation for char combustion, J/mol
E2 Energy of activation for calcination, J/mol
fc Fraction of heat given to coal particle released due to coal combustion
FC Coal flow rate, kg/s
Fp-w View factor
fv, 0 Initial mass fraction of volatiles in coal particle
Hc, C Heat of coal combustion in calciner, J/kg
Hc, K Heat of coal combustion in kiln, J/kg
Hcalc Heat of calcination reaction in calciner, J/kg
v
Hi Heat of formation of species i, J/kg
HLoss, K Heat losses in the kiln, J/kg
HR, K Heat required for clinker reactions, J/kg
hc Heat transfer coefficient between coal particle and gas, W/m2K
hc, c Heat transfer coefficient between clinker and gas W/m2K
hc,L Heat transfer coefficient between solid particle and gas, W/m2K
hcyc Heat transfer coefficient between particle laden gas and
cyclone inner wall, W/m2 K
h Height of solid bed, m
kg Thermal conductivity of air/gas, W/mK
kr Thermal conductivity of refractory, W/mK
ksh Thermal conductivity of shell, W/mK
ks,c Rate constant of char combustion, kg/m2s kPa
kapp Rate constant for pseudo homogeneous reactions, s
keff Effective thermal conductivity of gas in freeboard region, W/m K
kS0 Arrhenius factor for calcination reactions,
r
ks Rate constant of calcination of calcium carbonate, mol/m2s-1
r t
ks Rate constant of calcination of calcium carbonate, mol/m2s-1
ks Thermal conductivity of clinker, W/mK
L Total height of cyclone, m
Lgcl Chord of the sector covered by the charge, m
Len Length of the kiln, m
Mc, C Mass of coal entering the calciner, kg/s
Mc, k Mass of coal entering the kiln, kg/s
Mg Mass of gas in cyclones, kg/s
Mg, K Mass flow rate of secondary air entering the kiln, kg/s
Mg, P Mass flow rate of gas entering the pre-heater, kg/s
Mg, S Mass flow rate of secondary air entering the kiln, kg/s
Mg, T Mass flow rate of tertiary air entering the calciner, kg/s
Ms Mass of solids in cyclones, kg/s
Mse Mass of solids entrained by gas in cyclones, kg/s
Ms,C Mass flow rate of solids leaving the calciner, kg/s
Ms,K Mass flow rate of clinker leaving the kiln, kg/s
vi
Ms,P Mass flow rate of solids entering the pre-heater, kg/s
MwCaCO3 Molecular weight of Calcium carbonate, kg/kmol
MwCaO Molecular weight of Calcium oxide, kg/kmol
Mwchar Molecular weight of carbon, kg/kmol
MwCO2 Molecular weight of Carbon dioxide, kg/kmol
MwO2 Molecular weight of oxygen, kg/kmol
Mwvol Molecular weight of volatile, kg/kmol
Mww Molecular weight of water, kg/kmol
ma Mass of air in cooler, kg/s
mAl2O3, i Mass of total aluminum oxide in solids in calciner, kg/s
mCO2, C Mass of carbon-dioxide produced in calciner due to calcination, kg/s
mCaCO3, i Mass of total calcium carbonate in solids in calciner, kg/s
mFe2O3, i Mass of total ferrous oxide in solids in calciner, kg/s
mSiO2, i Mass of total silicon dioxide in solids in calciner, kg/s
mg Mass of gas in calciner, kg
mgin Mass of air entering in calciner, kg/s
mgout Mass of air leaving calciner, kg/s
mg, K Mass of air leaving the kiln calciner, kg
mpc, 0 Initial mass of coal particle, kg/s.
mp, c Mass of coal particle, kg.
mp, cin Mass of coal particle entering calciner, kg
mp,cout Mass of coal particle leaving calciner, kg
mp, L Mass of solid particle, kg
mp,Lin Mass of solids entering calciner, kg
mp,Lout Mass of solids leaving calciner, kg
ms Mass of solids/clinker in cooler, kg
MW Molecular weight, kg/ kmole
Mj Stiochiometric coefficient of the base component
mL Fraction of liquid formed due to melting
c
Np Number of coal particles entering calciner per second
L
Np Number of solid particles entering calciner per second
NC No. of components in the bed
NR No. of reactions
P The percentage calcination occurring inside the calciner
vii
∆P Pressure drop across the cyclone, mm of H2O
Pr Prandtl number
po2 Partial pressure of oxygen in gas, kPa
pco2 Partial pressure of carbon-dioxide in gas, kPa
peq Equilibrium partial pressure for carbon-dioxide in gas, kPa
Pe Peclet number for axial flow of solids
Q' Heat gained by the bed due to heat transfer, kJ / m2 s
Q '' Heat lost by gas to bed and walls, kJ/m s

QCGW Convection gas to wall, J/s
QCOAT Heat transfer through coating, J/ s
QCGB Convection from gas to bed, J/ s
QCWB Conduction from wall to bed, J/ s
QREF Heat transfer through refractory, J/s
QRGB Radiation from gas to bed, J/s
QRGW Heat transfer through refractory, J/s
QRWB Radiation from wall to solids, J/s
QSHELL Heat transfer through steel material, J/s
QRAD Radiative heat transfer between shell and secondary shell, J/s
QCONV-SHELL Convective heat transfer from shell to cooling air, J/s
QRAD-SHELL Radiative heat transfer from shell to air, J/s
QCONV-SEC-SHELL Convective heat transfer from secondary shell to air, J/s
QCOND-SEC-SHELL Conductive heat transfer in secondary shell, J/s
Q RAD-SEC-SHELL Radiative heat transfer from secondary shell to air, J/s
QINSU Conductive heat transfer in insulation layer, J/s
ri Internal diameter of cyclone, m
r0 External diameter of cyclone, m
rp Radius of solid particle, m
rr Internal diameter of cyclone shell, m
Re Reynolds number
R Internal radius of the kiln, m
RCOMB combustion rate of coal per unit volume of the gas (kg/m3 s)
RCOMBG combustion rate of volatiles per unit volume of the gas (kg/m3 s)
Rc Non-dimensional form of radiative heat transfer coefficient.
viii
Ri Net Rate of formation of species i in the bed, kg/m3 s
Rp Particle radius, m
R0 Initial particle radius, m
RSHELL Radius of kiln shell, m
RSEC-SHELL Radius of secondary shell, m
Tf Average temperature of solids and air in cooler, K
Tg Temperature of gas, K
Tg, in Temperature of gas entering calciner, K
Tg, out Temperature of gas exiting calciner, K
Tg, K Temperature of gas leaving the kiln, K
Tg, S Temperature of secondary air, K
Tg, T Temperature of tertiary air, K
Tg, P Temperature of gas leaving the pre-heater, K
TL Temperature of solid particle in calciner, K
Tiw, i The internal wall temperature of the cyclone, K
Tow, i The external wall temperature of the cyclone, K
Tr, i The temperature of interface of refractory and shell in cyclone, K
Ts Temperature of solids/clinker in cooler, K
Ts, C Temperature of solids entering the kiln, K
Ts, R Temperature of solids exiting the cooler, K
T0 Ambient air temperature, K
TAIR Temperature of air in annulus of two shells, K
Tb Temperature of bed, K
TC Temperature of coal particle, K
Tcl Temperature of the bed, K
Tcoat Temperature at which coating attains maximum thickness
Te Kiln external temperature, K
Ti Steel/Refractory interface temperature, K
Tg Temperature of gas, K
TL Liquidus temperature, K
Tw Temperature of inner kiln wall, K
TS Solidus temperature, K
TSHELL Temperature of kiln shell, K
TSEC-SHELL Temperature of secondary shell, K
ix
u0 Inlet gas velocity in cyclone, m/s
us, x Grate speed in x direction, m/s
us, y Grate speed in y direction, m/s
ug, x Air velocity in x direction, m/s
ug, y Air velocity in y direction, m/s
uc Velocity of coal particle, m/s
uc0 Initial velocity of coal particle, m/s
UCONV Convective heat transfer coefficient, W/m2 K
Vcl Velocity of the charge, m/s
Vreact Volume of reactor, m3
x Axial distance in the kiln from solid entrance, m
xr thickness of refractory, m
xs thickness of shell, m
xAl2O3, C Mass Fraction of aluminum oxide entering kiln
xCaCO3, C Mass Fraction of calcium carbonate entering kiln
xCaO, C Mass Fraction of calcium oxide entering kiln
xFe2O3, C Mass Fraction of Ferrous oxide entering kiln
xSiO2, C Mass Fraction of silicon dioxide entering kiln
yc, c Mass fraction of char in coal particle in calciner
yc, cin Mass fraction of char entering in coal particle
yc, cout Mass fraction of char leaving in coal particle
yc, K Mass fraction of char in coal particle entering the calciner
yv, c Mass fraction of volatiles in coal particle
yO2 Mass fraction of oxygen in gas
yO2, in Mass fraction of oxygen entering calciner in gas
yO2,out Mass fraction of oxygen leaving calciner in gas
yCO2 Mass fraction of carbon-dioxide in gas
yCO2, in Mass fraction of carbon-dioxide entering calciner in gas
yCO2, out Mass fraction of carbon-dioxide leaving calciner in gas
yv Mass fraction of volatiles in gas
yw Mass fraction of water in gas
Yi Mass fractions of the ith species in the bed
z Axial distance in the kiln from burner end, m
Zij Stiochiometric Coefficients
x
Greek Letters
ρcl Bulk density of the bed, kg / m3
ρb Bulk density of solids, kg / m3
αg Absorptivity of gas
β Angle of repose, rad
γ Kiln tilt, rad
δ Characteristic constant for kiln burner
Γ Angle made by solid surface at the kiln center
λ Latent heat of melting, kJ/kg
φv Volumetric flow of solids, m3/s
εb Emissivity of bed
εg Emissivity of gas
εs Solid porosity
εsolid Volume fraction of solids in the freeboard region.
εw Emissivity of Kiln internal wall.
Ω View factor for radiation.
Chemical Species
C2S (2CaO.SiO2)
C3S (3CaO.SiO2)
C3A (3CaO.Al2O3)
C4AF (4CaO.Al2O3.Fe2O3)
C12A7 (12CaO.7Al2O3)
C2AS (2CaO.Al2O3.SiO2)
CS (CaO.SiO2)
C3S2 (3CaO.2SiO2)
CS2 (CaO.2SiO2)
CF (CaO.Fe2O3)
C2F (2CaO. Fe2O3)
xi
1. Introduction
Abstract
In this chapter, an introduction to rotary cement kilns is presented. The motivation for
undertaking the present research is discussed. The methodology adopted is presented. The
overall organization of thesis is outlined.
1
1.1 Introduction
The manufacture of clinker formation in cement industry has been a focus of considerable
attention worldwide because of the high energy usage and high environmental impact of the
process. Typically, cement clinkers are produced by burning calcineous raw meal in a high
temperature environment. A schematic of a clinker manufacture is shown in Figure 1.1.
Calcineous Exhaust to atmosphere

Raw meal
Pre-heater Hot gases

Assembly to pre-heater
Coal
Tertiary Air
Calciner Kiln Exhaust

Pre-heated
Raw meal
Secondary Air
Primary Air
Kiln Coal
Cooler
Cooler clinker
Air to cooler
Figure 1.1: Schematic of clinker manufacture
The raw meal (consisting of predetermined quantities of CaCO3, SiO2, Al2O3 and Fe2O3) is
passed sequentially through pre-heater, calciner, kiln and cooler to form cement clinkers. In
pre-heater section the raw meal is preheated to calcination temperature via hot gases coming
from calciner. In calciner, calcination reaction occurs. The energy required for endothermic
calcination reaction is provided by combusting a suitable fuel. In most cases, coal is used to
provide the required energy, especially in India. The calciner is supplied with tertiary air from
the cooler and air coming out of kiln exhaust. The former is to supply sufficient O2 for coal
combustion and later to utilize the heat of kiln gases to enhance rate of heat transfer. The hot
gases from calciner are sent to pre-heater assembly for preheating the solids. The partially
2
calcined solids from the calciner are fed to slowly to the rotary kiln. In rotary kiln remaining
calcination and other clinkerization reactions occur (formation of C2S, C3S, C3A, C4AF). The
energy required for endothermic clinker reactions is provided by combusting coal in the kiln.
The pulverized coal along with the preheated air (secondary air) is fed to the kiln in a counter
current mode with respect to solids. Part of the solids melts in the kiln. The melt formation
causes an internal coating on the kiln refractories. The hot clinkers are discharged from kiln
to clinker cooler and hot gases from kiln exhaust are sent to the calciner. In clinker cooler a
part of energy of solids is recovered back by heat exchange with air. The air heated in the
coolers is passed to the kiln and calciner as secondary and tertiary air respectively. A small
part of air may be vented if required.
The ability to provide a high temperature environment, appropriate residence time for
clinkinerization reactions and handle solids with large size distribution makes rotary kilns the
most suitable reactor for clinker formation. Since the important reactions involved in clinker
formation occur in rotary kiln, performance of rotary cement kiln controls the quality of the
product and therefore the performance of the overall plant. Moreover, rotary kilns consume a
major portion of total energy supplied to the entire plant and are the main sources of CO2
emission in a cement industry. However, in spite of being a key unit and commonly practised
for decades in cement industry, realistic computational models to simulate cement kilns are
not readily available. A detailed understanding about cement kiln intricacies is still missing.
This research work was therefore initiated towards developing comprehensive computational
models to gain an understanding for rotary cement kilns and use this understanding for
possible performance enhancement. In the next section of this chapter, a brief discussion
about rotary cement kilns and key issues in cement kilns is presented. This discussion will
provide a background of challenges involved in developing realistic computational models
for rotary cement kilns. The motivation of the present research is presented subsequently.
Thereafter the methodology adopted in this work is discussed followed by outline of the
thesis.
1.2 Rotary Cement Kilns

Rotary kilns are commonly used in cement industry to convert calcineous raw meal to cement
clinkers. Typically rotary kilns used in cement industry are very long cylinders (generally
>50 m), which are slightly inclined to facilitate forward moment of the solids. The kilns are
3
revolved at pre-decided RPM to ensure uniform mixing of products. Schematic of a rotary
cement kiln is shown in Figure 1.2.
Coating
Rotating kiln
Exhaust gases
Secondary Air
Entrainment Radiation
Coal +
Primary Air
Melt
Partially Calcined Clinker
Raw Meal
Clinker Reactions
Flame
Figure 1.2: Schematic of rotary cement kiln
Partially calcined feed from calciner is fed slowly at velocity of ~ 0.05 m/s from one end. The
solid charge moves in a complex manner in the kiln due to a combined translational and
rotational motion. The solids in cement kilns are generally filled up to 10-15 % of fill to
ensure a uniform mixing. This part forms the bed region of the kilns. In the remaining part
called as freeboard region of the kiln, hot air is passed in a counter current mode with respect
to solids. Primary air and fuel are supplied in appropriate quantities through burner nozzles
with high velocities along with secondary air and swirling air to produce a stable flame in the
freeboard region. The flame is result of several number of combustion reactions involving
large number of components. The flame in the freeboard region is the direct source of heat
for endothermic reactions occurring in the bed region. The heat transfer in rotary kiln occurs
by various modes, which are described below
Radiative heat transfer between hot gases and the bed.
Radiative heat transfer between kiln refractory lining and the bed.
Convective heat transfer between hot gases and the bed.
Conductive heat transfer between kiln refractory lining and the bed.
Energy losses from shell via radiation and convection.
The partially calcined raw meal, which enters the kiln, utilizes the energy given by hot gases
to undergo various clinkerization reactions. In the initial part of kiln remaining calcination is
4
completed. This is followed by solid-solid reactions (formation of C2S, C3A, C4AF). A Part
of the solids melt in the kiln. One of the key reactions in clinker formation (formation of C3S)
occurs in the melt phase. The presence of melt causes a coating layer on the kiln refractories.
Formation of coating is considered to be beneficial from refractory point of view. The
location coating formation in the kilns govern the radiative heat losses from the shell.
Counter current flow of gas entrains solid particles from bed in the free board region. Such
entrainment enhances rates of radiative heat transfer by increasing effective emissivity and
conductivity. Thus in a rotary cement kiln, complex turbulent flow, heat transfer and
reactions occur simultaneously which involves multiple phases, vast number of species and
significantly different time and length scales.
1.3 Motivation of the Present Work and Methodology Adopted

The above discussion clearly indicates the complexity of physics involved during clinker
formation in cement kilns. For developing adequately accurate computational models for
cement kilns it is essential to consider the flow of solids in the bed and freeboard region,
clinkerization reactions in the bed region, coal combustion and gas phase combustion along
with radiative heat transfer. Other than this it is also essential to develop sub-models for
predicting melting of solids i.e. amount of melt formed in the burning zone, energy required
for solid melting, location of coating formation, influence of solid entrainment on the heat
transfer, etc. Moreover, there are numerical issues involved in cement kilns modeling as
discussed in the following. The physical time scales of gas and solids in cement kilns are
significantly different. The typical velocities of gas phase in the free board region is about
~10 m/s while typical velocities of solid particles in the bed region is of the order of a
typically about 0.05 m/s. In addition to the time scales of the free board and bed region,
actual chemical reactions occur at the scale of the particles, which is much smaller than
length scale of a kiln. Therefore incorporating various phenomena occurring simultaneously
in a reliable framework to predict the kiln behaviour accurately is a highly challenging task.
A detailed literature review on modeling attempts for cement kilns (as discussed in Chapter 2
later) indicates that none of the published models take into account all key factors (described
above) simultaneously. The motivation of the present work was therefore to fill this gap and
to develop comprehensive computational tools, which will capture most of the underlying
physics in cement kilns as shown in Figure 1.3. The objective of this research work was to
develop realistic computational models which consider simultaneous interaction of flow, heat
5
transfer and reactions occurring in cement kilns in a single framework and use these models
subsequently to identify the scope of possible performance enhancement. The methodology
adopted to achieve this goal is discussed in the next section.
Solid/solid; solid/liquid
reactions
Clinker formation
Granular Coal
flow models combustion/
(motion of turbulence/
solids in radiation
kiln) models
Figure 1.3: Integrated modeling approach for rotary cement kilns
Cement kilns are complex multiphase reactors where in different processes have significantly
different spatio-temporal scales as explained in previous paragraph. Hence, capturing all the
relevant processes in a single computational model, if not impossible, is an extremely
difficult task. Therefore, computational models were developed using different approaches
(reaction engineering based models and computational fluid dynamics based models), to
capture relevant physics which is explained further down in this section. These models have
evolved as different chapters of this thesis, which is explained further in Section 1.5 of the
thesis. Nevertheless, the models developed in different chapters contribute to the model
toolkit which is directed towards a single goal of developing computational models to capture
behaviour of rotary cement kilns and equipments associated with it. The approach adopted
here is pictographically demonstrated in Figure 1.4
The reaction engineering model was developed to gain an overall understanding of the kiln
behavior. This model was formulated based on one-dimensional framework for bed and
freeboard regions to allow effective simulations of overall kiln performances. Such reaction
engineering models have shown to complement more detailed CFD by models providing
basis for more rigorous CFD simulations of small number of promising cases (see Ranade,
6
V.V., 2002). The idea in developing reaction-engineering model was to develop a predictive
tool for cement kilns which would require order of magnitude lower computational resources
but still provide realistic simulations.
Complementary
CFD Model for cement kiln
1d Model for Kiln
Reaction Engineering Model

Pseudo-homogeneous bed CFD Model
To account
region/ variation of bed Coupling of bed and freeboard
for coupling
height/ plug flow/ 1D coal region, Eulerian-Lagrangian
with associated
combustion model with approach for the freeboard,
equipment
inputs from CFD simulations, Eulerian approach for the bed,
solids melting & coating details of burner configuration,
formation turbulence, radiation
RoCKS
Integrated Reaction
• Develop overall understanding • Develop comprehensive CFD tools Models for
Engg. model for Pre-heater
• Computationally cheap • Capture complex 3D flow solid flow in kiln
Calciner, Kiln &
Cooler • Capture performance with • Flame characteristics/radiation
reasonable accuracy • Establish effects of operating
parameters on kiln performance
Figure 1.4: Multi-layer modeling methodology for cement kilns
It is important to note that the utilization of stand alone model for the kiln would be rather
restricted since the pre-heater, calciner, kiln and cooler in clinker manufacturing are strongly
coupled to each other. Therefore reaction engineering computational models were developed
for pre-heater, calciner and clinker cooler. The kiln model was coupled with these models to
develop an integrated simulator for cement industry. The objective of developing integrated
simulator was to capture influence of key design and operating parameters on overall
performance (energy consumption per unit weight of product, production rate per unit
volume of kiln and so on) in clinker manufacturing.
The developed one dimensional reaction engineering model due to its inherent 1D nature will
not able to capture influence of burner design and key operating parameters like ratio of swirl
or axial air etc. on performance of cement kilns. Hence, it is essential to have a model, which
would provide quantitative guidelines to understand effect of burner operations on the
performance of cement kilns. CFD based models were therefore developed to achieve this.
Initially some attempts were made to develop CFD models for simulating solids flow in
transverse section of rotary kilns. The motivation of this work was to explore possibility of
using these models in CFD model for the cement kiln. However, these attempts were not very
7
successful. The CFD model for rotary cement kiln was therefore formulated assuming solids
as pseudo fluids.
1.4 Organization of Thesis

In this thesis work, computational models for rotary cement kilns and associated equipments
are developed and presented. These models have evolved as different chapters of this thesis
in three sections. The first section (Chapter 2) covers literature information about cement kiln
modeling. Cement kiln is a complex system wherein several processes occur simultaneously.
It is important to identify and adequately capture these key processes. In Chapter 2, these key
processes are identified. The challenges in modeling cement kilns are clearly outlined here
and status of the published models for rotary cement kilns is discussed. The remaining of
thesis is organized as shown in Figure 1.5.
Chapter 2 (Section 1)
Detailed literature review
Modeling of Rotary Cement Kilns
Reaction engineering Computational fluid dynamics

based models for cement based models for motion
kiln and associated of solids in rotary kilns/
equipments rotary cement kilns
Chapter 3 and Chapter 4 Chapter 5 and Chapter 6

(Section 2) (Section 3)
Chapter 7 (Conclusions)
Figure 1.5: Organization of the thesis
The second section presents reaction engineering based models for rotary cement kilns and
associated equipments. This section comprises Chapter 3 and Chapter 4 of the thesis.
Chapter 3 presents details of reaction-engineering models developed in this work. The
computational model presented in Chapter 3 was coupled with reaction engineering models
developed for pre-heater, calciner and cooler to develop integrated simulator for cement
8
industry in Chapter 4. The FORTRAN programs implementing the solution of integrated
simulator were modified to develop dynamically linked libraries (DLL) and interfaced with
Visual Basic to develop a simple graphical user interface based software called RoCKS
(Rotary cement kiln simulator) for cement industry. This is also discussed in Chapter 4.
In the third section of thesis (Chapter 5 and Chapter 6) computational fluid dynamics (CFD)
based models developed for rotary cement kilns are presented. Chapter 5 presents CFD
models developed for motion of solids and heat transfer in rotating cylinders. Chapter 6
reports development of comprehensive CFD model for rotary cement kilns. Finally the
conclusions of this research are outlined in Chapter 7.
9
2. Key Issues in Modeling Rotary Cement Kilns
Abstract
The aim of this chapter is to identify the key issues that need to be considered while
developing computational models for cement kilns. The important key issues viz. flow of
solids in bed and freeboard region, height variation of bed in the kiln, reactions occurring in
bed and freeboard region, melting of solids, coating formation are discussed one by one in
detail in this chapter. The modeling efforts to capture these issues in published literature with
their merits and limitations are discussed. The performance models for rotary cement kilns
are presented.
10
2.1 Introduction
Various processes occurring in rotary cement kilns need to be adequately considered while
developing its mathematical model. The key issues governing the performance of rotary
cement kilns are shown schematically in Figure 2.1. A partially calcined raw meal enters the
kiln with a certain flow rate. It is important to develop adequate models to estimate average
residence time of solids and variation of bed height within the kiln as a function of solids
flow rate, kiln rotational speed, tilt angle and so on. Several chemical reactions take place in
the solid bed. Calcination reaction liberates carbon dioxide and reduces solids mass flow rate.
Part of the solids melts in the kiln. The melt formation causes an internal coating on kiln
refractories. Energy required for clinkerization reactions and melt formation is provided by
the hot free board gases. Counter current flow of gas entrains solid particles in the free board
region. Such entrainment enhances rates of radiative heat transfer by increasing effective
emissivity and conductivity.
Coating
Dusty Gas
Entrainment due to counter current gas Gas Inlet
Raw Meal Clinker

Heat received by bed Melt
(Radiation)
Coating
Decomposition and Solid Solid Solid Liquid

Reactions Reactions
Figure 2.1: Key issues in modeling rotary cement kilns
Based on this brief description, the key processes occurring in cement kilns can broadly
classified as (See Figure 2.2)
1. Flow
2. Clinkerization reactions
3. Heat transfer.
11
Other than these, there are few other issues like melting of solids, coating formation on kiln
inner walls that occur in cement kilns during clinker formation. To develop realistic models
for cement kilns it is essential include all these physical processes simultaneously. We
discuss each of the issue and review the previous work related to it in following.
Reactions
Endothermic/Exothermic
Clinker Reactions
Coal Combustion/turbulent gas
Phase combustion
Flow Heat Transfer

Solids - Axial/Transverse Radiation/Convection/Conduction
Gas – Turbulent swirling flows Gas/Solids/Walls
Entrainment
Melting/Coating Formation
Figure 2.2: Schematic of various processes occurring in cement kilns
2.2 Flow of Solids in Rotary Cement Kilns

It is essential to accurately capture flow of solids in bed region while developing cement kiln
models. Solids flow in a complicated path in cement kiln having an axial component (due to
gravity) as well as radial component (due to rotation of kiln walls). It is essential to
understand motion of solids in both axial and transverse direction in cement kiln to capture
the residence time and mixing of solids in the kiln. We discuss about the same in this section.
2.2.1 Flow of solids in axial direction

The motion of solids in axial direction determines the solid residence time inside the kiln,
which directly affects the clinker composition. Axial motion of solids in the kiln is dependent
12
on the mass flux of solids, speed of rotation, height of solid bed (percentage fill), particle
size, angle of repose and kiln tilt. Axial velocity of solids varies along the kiln length causing
variation in height of the solid bed. Though reasonable models to simulate axial velocity of
solids in kiln are available, most of the models for cement kilns assume a constant bed height
(and subsequently constant axial velocity) within the kiln (Spang, 1972; Mastorakos et al.,
1999). Assuming a constant bed height will fix the residence time of solids in the kiln for any
set of operating conditions. Therefore, such models will not be able to predict the influence of
key design and operating parameters of kiln on its performance and their utility will be rather
restricted. It is therefore essential to use accurate models for axial motion of solids while
computational model for rotary cement kilns.
Attempts of experiments and computational models for understanding motion of solids in

rotating kilns in axial direction can be found in plenty (Friedman and Marshall, 1949,
Saemen, 1951; Kramers and Crookewit, 1952; Perron and Bui, 1990; Spurling et al., 2001).
Lebas et al. (1995) have published experimental data of residence time, particle movement
and bed depth profile in rotary kilns for a 0.6 m diameter kiln. Recently, Spurling et al.,
(2001) have also reported height variation of solid bed in a kiln of 0.1 m diameter. The
Kramer’s model (Kramers and Croockewit, 1952) which relates volumetric flow rate of
solids, φv, with kiln tilt (γ, radian), angle of repose (β, radian), radius of kiln (R, m), rotational
speed of kiln (n) and height of solids (h) was shown to predict both the set of experimental
data with reasonable accuracy as compared to other models published for rotary kilns (Lebas
et al., 1995). The description of Kramer’s model is discussed in Chapter 3 later and is
therefore not presented in this section. We have validated Kramer’s model with published
experimental data. The results of these simulations are shown in Figure 2.3, Figure 2.4 and
Figure 2.5. Figure 2.3 and Figure 2.4 shows comparison of Kramer’s model with
experimental data of Lebas et al. (1995). It can be seen from these figures that the model is
able to predict the data reasonably well for changes in kiln RPM and solid throughput to the
kiln. Same results were observed with published data of Spurling et al. (2001) (See Figure
2.5). Hence, we have used Kramer’s model for simulating bed height in this work.
13
0.2
Qs = 246 kg/hr
0.18 Qs = 350 kg/hr
Qs = 480 kg/hr
Kramers Model
0.16
Kramers Model
Kramers Model
0.14
0.12
Heigth (m)
0.1
0.08
0.06
0.04
0.02
0
0 1 2 3 4 5 6
Normalized Kiln Length (-) (m)
Figure 2.3: Comparison of model of Kramers and Croockewit (1952) with experimental
data of Lebas et al. (1995) for varying flow rate
14
0.18
2 RPM
0.16 Kramers Model
3 RPM
Kramers Model
0.14 4 RPM
Kramers Model
0.12
0.10
Height (-)
0.08
0.06
0.04
0.02
0.00
0 1 2 3 4 5 6
Normalised Kiln Length (-)
(m)
data of Lebas et al. (1995) for varying rotational speed
15
0.03
Qs = 0.625 g/s
Kramers Model
0.025 Qs = 2.45 g/s
Height variation along kilon Length (m)
Kramers Model
0.02
0.015
0.01
0.005
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalised Kiln Length (-)
data of Spurling et al. (2001) for varying flow rate
16
Other than motion of solids in axial direction, it is also essential to model solids mixing in
axial direction. Several methods have been used to analyze mixing of solids in rotating kilns
such as: 1) Finite stage transport 2) Monte Carlo simulations 3) Discrete element modeling
and 4) Dispersion model (Sherritt et al. 2003). Axial dispersion model is the most common
method of modeling axial mixing in both continuous and batch operation of rotary kilns and
will be used to study axial mixing in this work. In this model, the particles are treated as
continuum and the solids mixing in the axial direction is simulated using the following one-
dimensional equation:
∂c ∂ 2C ∂c
= DZ 2 − u Z (2.1)
∂t ∂t ∂z
where Dz is the axial dispersion coefficient. Higher magnitude of axial dispersion coefficient
indicates more mixing of solids.
Previous work on quantifying axial mixing in rotary cement kilns using axial dispersion
model were reviewed in detail by Sherritt et al. (2003). Sherritt et al. (2003) has reviewed
previous experiments with batch as well as continuous operation of kilns. They have analysed
a total of 179 data points from the literature and proposed design correlations for slumping,
rolling/cascading and cataracting bed behaviours. The following correlation was proposed by
Sherritt et al. (2003) to estimate axial dispersion coefficient:
a
⎛ n ⎞
DZ = k ⎜⎜ ⎟⎟ (2 R) b d p c X d (2.2)
⎝ NC ⎠
where Dz is the axial dispersion coefficient, k is regression parameter (Range – ln k = -0.75 ±
0.53 for rolling and cascading mode). Other parameters are reported as:
a = 0.44 ± 0.14
b =1.29 ± 0.16
c =0.35 ± 0.06
d = -0.55 ± 0.16
n – rotational speed.
60 g
Nc – critical rotation speed =
2π R
R – radius of drum.
dp – diameter of particle.
17
X – percentage fill.
We have performed several numerical experiments to get range of values of dispersion
coefficients from Equation 2.2. The data obtained from an industrial kiln (R =1.7 m) was
used to estimate axial dispersion coefficient for different operating conditions (percentage fill
– 0.1 to 0.5; rotational speed – 1 to 6 rpm; particle size – 1 µm to 10 mm) as shown in Table
2.1, Table 2.2 and Table 2.3.
Table 2.1 Variation of axial dispersion coefficient/Peclet number with percentage fill
Percentage fill DZ (m2/s) × 105 Pe × 10-6
0.1 18.83 5.3
0.2 12.86 7.8
0.3 10.29 9.7
0.4 8.79 11.4
0.5 7.77 12.9
Table 2.2 Variation of axial dispersion coefficient/Peclet number with rotation speed
Rotational speed DZ (m2/s) × 104 Pe × 10-6
(rpm)
2 1.01 10.
3 1.21 8.2
4 1.37 7.3
5 1.51 6.6
6 1.64 6.1
Table 2.3 Variation of axial dispersion coefficient/Peclet number with particle size
Particle size (m) × 104 DZ (m2/s) × 104 Pe × 10-6
0.01 0.31 31.8
0.1 0.7 14.2
1.0 0.58 6.3
10.0 3.53 2.8
100.0 7.90 1.3
18
0.35
0.3
plug
pe = 1000
pe = 100
0.25 pe = 10
Mas s Frac tion Ca C
3
0.2
0.15
0.1
0.05
0
0 0.2 0.4 0.6 0.8 1
Normalis ed Kiln Length (-)
Figure 2.6 Effect of axial dispersion coefficient on the kiln performance
This study indicated that for wide range of operating conditions covered in Table 2.1 to 2.3,
the values of Peclet number (uzL/DZ) were always greater 104. Thus even for a very wide
spectrum of kiln operations the value of Peclet number was always over 104. In order to
quantify influence of axial mixing on kiln performance, we have carried simulations were
carried out for different values of Peclet number for a kiln wherein transverse gradients were
neglected. For Peclet numbers higher than 1000, predicted results were almost
indistinguishable from those obtained from considering the plug flow. Though usual Peclet
numbers in the kilns are very high as discussed above, simulations were carried out at much
lower values to quantify the sensitivity. The results of these are shown in Figure 2.6.
For these simulations effect of varying Peclet number on CaCO3 concentration in the kiln was
monitored. These simulation results indicate that solids mixing in axial direction can be
safely neglected while developing computational models for cement kilns.
2.2.2 Flow of solids in transverse direction

The motion of solids in transverse direction is important from the product uniformity point of
view. With increasing rotation speed six flow regimes in transverse section of rotating kilns
19
are identified (Heinen et al., 1983). These flow regimes viz. slipping, slumping, rolling,
cascading, cataracting and centrifuging are shown in Figure 2.7.
Figure 2.7 Flow regimes in transverse section
The nature of solids flow in transverse section can change significantly with particle
characteristic, kiln size, rotation speed and bed fill. The surface renewal rate, which
determines heat transfer, mass transfer and mixing in bed depends on the prevailing flow
regime. Of all the regimes mentioned above, rolling mode offers maximum surface renewal
rate with good heat transfer, mass transfer and reactions. Therefore it is the most commonly
used mode in practice for cement kilns. The rolling mode is characterized by two distinct
zones as shown in Figure 2.8.
20
Passive Layer
Active Layer
Figure 2.8: Flow of solids in rolling regime
A thinner active layer which is formed as the solids flow down the sloping upper bed surface
and the much thicker passive layer (also called as plug flow region). In the passive region, the
granular material moves as a rigid body and particle mixing is negligible. Particle mixing
mainly occurs in the active region. The rate of such a mixing determines the surface renewal
rate and hence the rates of bed-freeboard heat and mass transfer, as well as chemical
reactions. This in turn influences mixing of solids in transverse section. Therefore it is
essential to understand behavior of solids in transverse section of cement kilns.
Motion of solids in transverse section of rotating cylinders depends on kiln rotational speed,
percentage fill and rheology of particles. The development of computational model for feed
bed that includes all the above mentioned interactions is a topic of intense research itself. The
constitutive equations for granular flows as yet are far from being established. The
approaches for developing computational models for motion of solids in transverse section
have been mainly based on Eulerial-Eulerian approach. Boateng and Barr (1996) developed a
model, which employed boundary layer approach in deriving governing equations. They
presented their integro-differential set of equations, which contained granular temperature
term. Khakhar et al., (1997) have presented an excellent work on motion and mixing of solids
in transverse section of rotating drums. They have used a macroscopic balance approach to
obtain velocity and layer thickness for the flow in the active layer. Ding et al., (2001)
developed a more sophisticated model considering a quasi-static contribution, which becomes
important at low rotational speed. This term was not considered in Boateng’s model. The
21
above-mentioned models use a polynomial function for the velocity profile in the active layer
to get the solution to the integro-differential equations. This velocity profile is mostly
obtained from experiments. The models assume surface of solids to be flat which is not really
the case especially near the walls. Moreover, most of this work is for solids with high fill
ratio’s or low rotational speeds. Further, these models are restricted to the rolling mode of
operating regime. Therefore, more generalized models to capture transverse flow of solids in
rotating cylinders are required to be developed which may account for different particle sizes,
shapes and properties at higher rotational speeds.
2.3 Clinkerization Reactions

Several solid-solid and liquid-solid reactions occur in the bed region of rotary cement kilns.
The chemical reactions that occur in the kiln are described in detail by Hewlett, (1998). The
most important oxides that participate in the reactions are CaO, SiO2, Al2O3 and Fe2O3. The
reactions occur in a particular sequence along the kiln length. Initially (~ 700–900 0C)
calcination as well as an initial combination of alumina, ferric oxide and silica with lime
takes place. Between 900 and 1200 0C belite [C2S, i.e. (2CaO,SiO2)] is formed. Above 1350
0
C, a liquid phase appears and this promotes the reaction between belite and free lime to form
alite, [C3S i.e. (3CaO,SiO2)]. During the cooling stage, the molten phase forms tri-calcium
aluminate [C3A i.e. 3CaO,Al2O3] and a ferrite phase, [C4AF i.e. 4CaO, Al2O3, Fe2O3]. If the
cooling is slow, alite may decompose and appear as secondary belite and free lime. In the
burning zone approximately 25 % of the material remains in the liquid state. At this stage the
materials become ball of fused material ranging in particle size from big particles to as fine as
dust (~ 50 µm – 5 mm). These materials are known as cement clinker. The parameters that
control these overlapping processes are the mineralogy of the raw materials, the fineness of
the raw materials, and the homogeneity of the raw mix.
Kinetics of cement clinker formation is not well understood in all details. Decomposition of
solid-solid reaction between and other oxides like and solid-liquid reaction between and di-
calcium silicate (formed in solid-solid reaction between CaO and are the key reactions and
play most important role in the context of cement clinker formation in kiln. The main
chemical reactions occurring in the kiln during clinker formation are given below:
22
Dissociation of CaCO3 (Solid Decomposition)
CaCO3 Î CaO + CO2 ∆HR = 43.74 kcal/mol (2.3)
Combination of lime and clay (Solid-Solid Reaction)
CaO + SiO2 Î C2S ∆HR = -33.24 kcal/mol (2.4)
3CaO + Al2O3 Î C3A ∆HR = -5.8 kcal/mol (2.5)
4CaO+Al2O3 + Fe2O3 Î C4AF ∆HR = -12 kcal/mol (2.6)
Solid liquid reaction (Liquid Phase Sintering)
CaO + C2S Î C3S ∆HR = 0.47 kcal/mol (2.7)
Reaction 1 (Equation 2.3, calcination reaction) is a decomposition reaction. Reactions 2, 3

and 4 (2.4, 2.5 and 2.6 respectively) are the solid-solid reactions. Reaction 5 is the solid-
liquid reaction. While formation of some other compounds like C12A7, C2AS, CS, C3S2, CS2,
CF, C2F etc. has been reported within the kiln (Hewlett, 1998), these components are
generally present in insignificant amount and are usually neglected. In general, only the
major reactions taking place in clinker formation are considered for modeling rotary cement
kilns (Spang, 1972, Mastorakos et al., 1999). In this work therefore only five major reactions
to occur during clinker formation are considered. Modeling clinkerization reactions and
estimating relevant kinetics for these reactions is not a straightforward task. In this section the
models available for the three major reaction mechanisms occurring in the clinker formation
are described briefly.
Calcination reaction (Equation 2.3) is a decomposition reaction and has been studied
extensively. The overall reaction rate is mainly controlled for the following three rate
processes, 1) heat transfer (to the surface and then to the reaction zone) through the particle
2) mass transfer (both internal and external to the surface) and 3) chemical reaction at the
interface of undecomposed particle and product. Relative rates of these three transport
processes depend on particle size and operating conditions. Several models such as shrinking
core model, grain model, uniform reaction model, nucleation and growth model and reacting
particle model have been proposed for modeling decomposition of Calcium Carbonate (see
for example, Ingraham and Marier, 1963; Hills, 1968; Khinast et al., 1996; Irfan and Dogu,
2001; Martins, 2002). The shrinking core models were shown to be applicable to wide range
of limestone species under conditions similar to that of kiln.
23
The solid-solid reactions occur after decomposition as described by Equation (2.3), Equation
(2.4) and Equation (2.5). Modeling solid-solid reactions is not very straight forward. As a
general rule, solid-solid reactions demonstrate profound multi-step characteristics. This can
involve several processes with different activation energies and the mechanisms. Several
attempt have been made in literature for predicting the formation kinetics of different calcium
silicate, calcium aluminate and aluminoferrite phases. These phases formed due to solid solid
reaction between lime and other oxides like silica, alumina and iron oxides. Many theoretical
models have been derived over the past 70 years for the solid-state reaction of powders, but
these models, based on various conceptualisations of the physicochemical phenomena
involved, are not always successful all the way up to 100% reaction (Hao and Tanaka, 1990).
The rate of the reaction of the solids depends on particle size; size distribution and previous
history of the reactants themselves and these dependencies are so complex that kinetic
measurements are difficult to interpret (Beretka and Brown, 1983).
The solid liquid reaction takes place mostly in the burner zone where the temperature of the
0
charge is greater than 1350 C. The most important and virtually the slowest step in the
clinker formation is the formation of tricalcium silicate (C3S) from uncombined lime (CaO)
and dicalcium silicate (C2S) in liquid phase (Johansen, 1973). This process is most important
because C3S is the most powerful hydraulic constituent of the cement and a low content of
unreacted lime is required for a stable cement mortar. In fact the content of C3S is the
indicator of the extent of cement formation reaction. The original inhomogeneities of raw mix
will result in regions of different composition which makes the CaO diffuse through the
liquid reaction layer surrounded it. Assuming the unchanged contact face original radius
during the reaction and applying the spherical diffusion field Johansen (1973) developed a
model equation of fraction of reacted CaO in time. Using the model, the maximum size of the
particle which can react fully within a given time can be estimated, as well as, for a given
particle size distribution, the free content of the clinker.
Reviewing literature of clinkerization kinetics survey suggests that calcination reaction

(Equation 2.3) has been studied extensively. Unlike calcination, other clinkerization (like
formation of C2S, C3A, C4AF and C3S) are not extensively studied. Particle level models for
these reactions are not available. Most of the previous studies are restricted to characterizing
diffusion (and reactions) of CaO in silica and other oxide species. Adequate information
about variation of diffusion coefficient with temperature and composition under conditions
24
prevailing in kiln is not available. These reactions are therefore usually modeled as pseudo-
homogeneous reactions (Spang, 1972, Mastorakos et al., 1999). Ignoring diffusion and
particle scale processes, simple Arrhenius type expressions were used for describing overall
reaction rates in this work.
2.4 Melt and Coating Formation

Another important aspect of cement kilns is the formation of melt within the kiln. In the
burning zone of the kiln the solids are exposed to extremely high temperatures (of the order
of 1700 K). As a result of this solids melt. Melt formation is essential for formation of C3S.
Despite this, not many of the published models for cement kilns account for formation of a
melt phase in cement kilns. Mastorakos et al. (1999) had accounted for melting in cement
kilns. However their model arbitrarily assumed a linear increase in the melt phase formation
till a certain maximum melt fraction (around 30 %). The melt fraction was assumed to remain
constant thereafter in the burning zone till the solids exit. The amount of melt formed
depends on the bed temperature profile in the kiln. As the melt moves towards exit of the kiln
it may re-solidify since the bed temperature decreases near the solids exit. However, none of
the previous models account for re-solidification of the melt.
Figure 2.9: Reported shell temperature measurements from Kolyfetis and Markatos,
(1996) [Kiln Length: 75 m, Inner diameter: 4.35 m]
25
As the melt phase is formed, it forms a coating over the refractories in the remaining part of
the kiln. The formation of coating is beneficial for the refractory life. The computational
models for cement kilns presented earlier, either consider a uniform coating throughout the
kiln (Mastorakos et al., 1999) or do not consider any coating formation in kiln (Spang, 1972).
Both the above assumptions are neither correct nor acceptable for realistic simulations.
Coating formation starts at some distance from the solids entry after the solids melt. Such a
non-uniform coating formation leads to discontinuity in kiln shell temperature with a sharp
fall (~ 200 oC) in shell temperature at the point of coating formation as shown in Figure 2.9
(reproduced from Kolyfetis and Markatos, 1996). Coating formation has a direct impact on
losses from kiln shell. Hence predicting accurate coating formation is essential for predicting
correct figures for energy consumption.
2.5 Heat Transfer

Heat exchange between free board and solid bed needs to be captured accurately since this
drives the chemical reactions in the bed. Other than heat transfer between bed and freeboard
air there is heat exchange between gas and kiln walls and bed and kiln walls. Heat transfer
occurs through all the three modes viz. conduction, convection and radiation. The various
modes by which heat transfer takes place in kiln are shown in Figure 2.10
QLOSS
QSTL
QREF
Steel QCGW
QRGW QRWB
QCGB
QRGB Γ
Bed
QCWB
Refractory
Figure 2.10: Heat transfer in rotary cement kilns
QRWB and QCWB are heat transfer rates due to radiation and conduction respectively between
kiln internal wall and bed. QRGB and QCGB are heat transfer rates due to radiation and
26
convection between gas and bed. QRGW and QCGW are heat transfer rates due to radiation and
convection between the kiln freeboard gas and internal wall. QREF and QSTL are heat transfer
due to conduction in refractory and steel shell respectively. QLOSS is loss of heat from steel
shell by radiation and convection. Capturing the heat transfer rates accurately is also
important to develop comprehensive cement kiln models. For operating conditions prevailing
in cement kilns, radiation dominates the heat transfer process (Boateng and Barr, 1996;
Mastorakos et al., 1999; Karki et al., 2000). Key parameters appearing in heat transfer
models are: emissivity of gas, absorption coefficient of gas, conductivity of gas & solid.
Estimation of these properties is not straightforward. Since the temperature range (minimum
and maximum) in the freeboard region varies significantly, the thermal properties of gas will
not be constant and a function of local temperature in the kilns. Moreover, because of the
counter current flow gas flow, solids will get entrained from bed to freeboard region. The
extent of solid entrainment will influence the effective emissivity as well as conductivity of
gas. Hence effective thermal properties of dust-laden gas need to be estimated to model heat
transfer in transverse direction accurately.
Due to extremely high temperatures encountered in the freeboard region of kiln radiation
dominates the heat transfer process. Characterizing radiation energy in the freeboard region
of kiln is a complex but crucial element in modeling of cement kilns. Eaton et al. (1999) have
given a brief and nice explanation of models for characterizing radiation in pulverized coal
combustors. The authors have also presented further in-depth review articles for the
interested readers. A basic integro-differential equation is well developed for steady state
radiative heat transfer and is given as
dI ( s, ω ) σ
ds
= −(κ + σ ) I ( p, ω ) + κI b +
4π ∫π I ( s,ω ) Φ dw
4
(2.8)
where I is the radiative intensity, p and ω are unit vectors in the directions of propagation, κ,
and σ are the local absorption and scattering coefficients, respectively, and Φ represents a
phase function used to characterize the nature of the scattering media. The term on the left
side of the equation represents the gradient of intensity in the specified direction, v. The three
terms on the right side of the equation represent the changes in intensity due to absorption
27
and out-scattering, emission and in-scattering respectively. The references describing various
approaches for derivation of equation can be found in Eaton et al. (1999).
The accuracy of the solution, however largely depends on the accurate knowledge of
radiative properties of gas. Key parameters appearing are: emissivity of gas, absorption
coefficient of gas, conductivity of gas & solid. Calculation of these properties require
information on temperatures, gas composition as well as particle concentration, composition,
shape and size distribution. Seluck et al. (2002) have presented estimation of thermal
properties for freeboard region of bubbling fluidized bed combustor. In this work we have
used volume fraction weighted averaged values to calculate effective values of thermal
properties for industrial cement kilns. However, from the present literature it seems that the
data presented for predicting thermal properties of particle-laden gas is very limited.
Knowing the radiative properties of combustion gases, the radiative equation needs to be
solved to quantify the radiation in the freeboard region. Since it is not possible to develop a
single solution method that is applicable to wide variety of systems (Eaton et al., 1999) and
therefore several solution methods have been developed to solve radiation equation (equation
2.8). The different methods are evolved in varying degree of assumption considering the
nature of the system, characteristic of the medium, degree of accuracy required, etc. The
major approaches can be classified as (Eaton et al., 1999)
• Statistical methods
• Zonal method
• Flux methods; including the discrete-ordinates approximation
• Moment methods
• Spherical harmonics approximation
• Hybrid methods.
Some of the above solution methods are also applied to estimate radiative heat transfer in
rotary kilns. The standard solution techniques are also available in commercial CFD codes
(Fluent 6.1.18). Multi flux radiation model is commonly used by lot of modelers. Karki et al.
(2000) have used six flux radiation model with isotropic scattering. In six flux model
radiation field is assumed to be represented by six fluxes at each point in the space. The flux
in the each direction is assumed to change according to local emission, absorption and
scattering. Guo et al. (2003), have used four heat flux radiation model. Marias (2003) have
28
used P-1 radiation model. Mastorakos et al. (1999) have used a statistical approach to
estimate radiative heat transfer in cement kilns. The radiation intensity of the gas was
calculated with Monte-Carlo method. Since the motivation of present work was not to study
radiative heat transfer in cement kilns, standard models available in commercial codes were
used to model radiation in cement kilns in this work.
2.6 Freeboard Region of Cement Kilns
Accurate modeling of freeboard region is a key in developing computational models for

rotary cement kilns. The main function of the freeboard region is to supply sufficient energy
to the solid bed for driving solid-solid and solid-liquid endothermic reactions for clinker
formation. This energy is obtained by burning fuels like coal, natural gas, etc. Rotary kilns
can be fired with different fuels. However, most of the industrial cement kilns are equipped
with coal-fired burners, especially in India. Primary air and fuel are supplied in appropriate
quantities through burner nozzles along with secondary air and swirling air to produce a
stable flame in the freeboard region. The burner configuration, swirl air and other operating
parameters influence the flame characteristics, heat transfer to the bed region and temperature
profiles in the bed and the freeboard regions. The flame supplies the required energy to the
solid bed for clinkerization. The heat transfer is mainly through radiative heat transfer as
discussed in the previous section.
The main processes that occur in freeboard region are pictorially represented as shown in
Figure 2.11. It is essential to include component models for the representation of turbulent
fluid flow and heat transfer, turbulent combustion, particle reactions simultaneously to
simulate freeboard region. The gaseous combustion can be either be reaction limited or
mixing limited or both depending on the kiln operating conditions. The burners of cement
kilns produce a complex flow. The velocity profiles of air are highly non-uniform especially
near the kiln burners. It is therefore essential to capture swirling turbulent flows produce by
kiln burners in the freeboard region. The burner configuration (for example, number of axial
air inlets, Location of burner, proportion of swirl to axial air ratio) influences the flow and in
turn flame characteristics of the freeboard region.
29
Figure 2.11 Various physico-chemical process occurring in kiln freeboard region
It is essential to properly simulate turbulent gas phase combustion and coal combustion in the
freeboard region of cement kilns. Modeling non-premixed turbulent combustion processes
requires an effective scheme for simultaneously modeling both mixing and reactions. Eaton
et al. (1999) have presented in detail about modeling gas phase combustion reactions in coal
combustors in their review. It is essential to calculate individual time scales or rates of mixing
and reactions in freeboard region. In employing simplifying assumptions to achieve feasible
solutions of turbulent gas phase combustion, three modeling approaches have been used.
These approaches are classified according to two hypothetical time scales associated with
chemical reactions in turbulent flow: the reaction time scale and the turbulent mixing time
scale. The reaction time scale is typical time required for the species of interest to react
completely. The turbulent mixing time scale is the typical time required for large scale
turbulent eddies to break up and to reduce to the scale where molecular interactions can
occur. In the first modeling approach, which is not expected to occur in industrial kiln,
reacting species are assumed to be premixed or the turbulent mixing scale is assumed to be
infinitely fast. Therefore in this case turbulent mixing can be ignored. In other modeling
approach, the reaction rate is assumed to be infinitely fast and combustion is assumed to be
mixing limited. In the third modeling approach, when both the rates are of same order
different reaction rates are calculated and smallest rate is assumed to be the governing rate.
Other than gas phase combustion coal combustion also occurs in the freeboard region. Coal
combustion is known to occur in four stages viz. 1) heat up, 2) devolatilisation of volatiles, 3)
30
combustion of volatiles and 4) char burnout. Heat up of coal is due to radiation and
convection from surrounding gas (secondary air). Heat up leads to release of volatiles, which
may burn if temperature is sufficiently high, and oxygen is available. The final stage of coal
combustion is char-oxidation. Char oxidation is a relatively slow process as compared to
other two steps and is a heterogeneous reaction. For char oxidation to take place oxygen has
to diffuse through the pores of charcoal and then reacts with carbon to form the products.
Depending upon the temperature conditions diffusion and/or chemical reaction may limit the
process. It is necessary to account for different phenomena and their corresponding rates
while developing the model for the free board region of cement kilns.
Several attempts of modeling these processes have been published. Modeling of heating up
phase of coal particle is fairly straightforward. Modeling of coal devolatilization is also fairly
well established. Most widely used approaches use correlations of volatile yield with particle
temperature or define devolatilization rates using single or two step Arrehenius schemes.
Modeling char combustion is however not straightforward. Char combustion is controlled by
a complex coupling between transport phenomena around and within the particle. The
problems associated with calculation of rate of char combustion are due to uncertainty in
morphology of devolatilized coal particle, exact composition of coal and effects of ash.
Computationally intensive network models, which are based on coal structural description,
are not yet well established. Therefore most of the models proposed for char combustion are
empirical in nature (Williams et al., 2002)
2.7 Summary of Key Issues
The various physical issues which were discussed in detail in this section are summarized
below.
• It is essential to simulate motion of solids/bed height in the bed region. Axial

dispersion of solids can be neglected while developing computational models for
rotary cement kilns.
• It is essential to account for motion of solids in transverse section. The prediction
of flow in active layer/passive layer, rate of surface renewal, needs to be
understood to simulate mixing/segregation in solid bed
31
• The melt formation in cement kilns and the re-solidification of melt near kiln exit
should be considered during model development. Accurate prediction of location
of coating formation is essential for proper estimation of heat losses in the kiln
and shell temperatures.
• The physical properties of particle laden gas should be considered while
developing models for capturing heat transfer in cement kilns. It is essential to
accurately capture heat exchange specifically radiation between solids, gas and
kiln walls.
• Freeboard region where turbulent fluid flow along with gas phase combustion and
coal combustion occurs needs to be simulated accurately.
2.8 Performance Models for Rotary Cement Kilns

Considering the key issues (and relevant previous work) discussed above, it is not surprising
that computational models are not readily available for simulating rotary cement kilns. It is
evident that in order to develop a reliable computational model it is required to incorporate all
the above-mentioned processes occurring in cement kilns. However, none of the published
models for cement kilns have done this. This is what we aim to do in this research work. The
performance models developed for cement kilns were mainly based on two approaches viz.
reaction engineering based models (incorporating material and energy balance but not
considering momentum balances) and CFD based models (incorporating all the three
fundamental balance equations). However, the computational models published for cement
kilns based on either of approaches are not very comprehensive. The lacunae of the published
models are discussed below to clearly bring the need of tractable but realistic computational
models for rotary cement kilns.
2.8.1. One dimensional models

Few one dimensional models were developed in about couple of decades for rotary kilns.
(For example, See Watkinson and Brimacombe, 1982 and references there in). However,
most of models were limited for studying heat transfer or flow in rotary kilns. Very few
models (Spang, 1972) were specifically proposed for cement kilns. The published 1D models
for cement kilns do not take into account simultaneously variation of bed height variation,
flow, heat transfer, melting, coating formation, and reactions in the bed as well as free board
(See for example, Spang, 1972; Mastorakos et al., 1999). Unless this is done, it is impossible
32
to capture influence of design (diameter, length, tilt angle) and operating parameters (RPM)
on kiln performance. Unless the kiln simulation model is able to capture influence of design
and operating parameters of kiln on the kiln performance, the utility of the model will be
rather restricted. None of the models for cement kilns consider melting of solids and re-
solidification of melt correctly (Mastorakos et al., 1999). Most of the models assume uniform
coating along the kiln length (Spang, 1972; Mastorakos et al., 1999). This can lead to
erroneous calculation of losses from kiln shell. Thus, more comprehensive modeling
framework is required to be built for rotary cement kilns.
2.8.2 CFD Models

Due to complexity of physics involved, occurrence of multiple phases with large number of
reactions in bed/freeboard regions and numerical issues, very few CFD models have been
published for rotary cement kilns (Kolyfetis and Markatos, 1996; Mastorakos et al., 1999;
Karki et al., 2000). Mastorakos et al. (1999) developed a computational fluid dynamics
(CFD) based model for cement kilns, which included combustion, radiative heat transfer,
conduction in the bed/ walls and chemical reactions. The bed and freeboard models were thus
treated as separate domains and coupling between them is handled explicitly. The geometry
of kiln was assumed to be axisymmetric in this work and therefore the boundary conditions
were applied only in an approximate manner. Moreover Mastorakos et al. (1999) assumed
coating formation throughout the kiln length. Karki et al. (2000) developed a 3D CFD based
model for simulating simultaneous combustion and heat transfer in cement kilns.
Karki et al. (2000) have used an effective thermal conductivity to define degree mixing in
bed region, developing a single computational model for simulating cement kilns. Different
values of effective thermal conductivities at different locations in the kiln were used.
However, there are no proper guidelines to choose proper effective thermal conductivity and
the values used are based on experience. Moreover Karki et al. (2000) have focused on fluid
dynamics and heat transfer and did not model clinkerization reactions. The model predictions
thus gave only qualitative agreement with industrial observations reasonable for global
quantities of interest. Kolyfetis and Markatos (1996) also focused on coal combustion and
heat transfer in freeboard region and did not account clinker reactions. Moreover, Kolyfetis
and Markatos (1996) did not consider any coating formation in their computational model.
From this literature review, it can be concluded that the computational models presented
earlier for rotary cement kilns are more or less approximate and not rigorous models.
33
Therefore, comprehensive model for cement kilns need to be developed which will account
for simultaneous flow, heat transfer and reactions occurring in rotary cement kilns. Such an
attempt is made in this work which will be present in remaining part of this thesis.
34
3. Reaction Engineering based Model
Abstract
In this chapter, a comprehensive reaction engineering based model to simulate complex
processes occurring in rotary cement kilns is presented. A modeling strategy comprising
three sub-models viz. model for simulating variation of bed height in the kiln, model for
simulating reactions and heat transfer in the bed region and model for simulating coal
combustion and heat transfer in the freeboard region was developed. Melting and formation
of coating within the kiln were accounted. Combustion of coal in the freeboard region was
modeled by accounting devolatilization, finite rate gas phase combustion and char reaction.
The simulated results were validated with the available data from 3 industrial kilns. The
model was then used to understand influence of various design and operating parameters on
kiln performance. Several ways of reducing energy consumption in kilns were then
computationally investigated. The model was also used to propose and to evaluate a practical
solution of using a secondary shell to reduce energy consumption in rotary cement kilns.
Simulation results indicate that varying kiln operating variables viz. solid flow rate or RPM
can result only in small changes in kiln energetics. Use of secondary shell over kiln and
energy recovery by passing air through the annular gap between the two, appears to be a
promising way for significant energy saving. The developed model and the presented results
will be useful for enhancing performance of rotary cement kilns.
Publications based on this work

• Mujumdar, K.S., Arora, A., and Ranade, V.V. (2006), “Modeling of Rotary Cement Kilns:
Applications to Reduction in Energy Consumption”. Ind. Eng. Chem. Res., 45 (7), 2315 -
2330, 2006
• Kaustubh S. Mujumdar and Vivek V. Ranade (2006), “Simulation of Rotary Cement Kilns
Using a One-dimensional Model”, Chemical Engineering Research and Design, 84, 165-
177
35
3.1 Introduction
Rotary kilns are commonly used in cement industry to produce cement clinkers. Most of the
clinkerization reactions occur in rotary kilns and therefore it governs overall performance of
clinker manufacture. However, computational models are not readily available for rotary
cement kilns in published literature. This can be attributed to involvement of complicated
physics involved in clinker manufacture as discussed in chapter 2. However, there is an urge
for developing comprehensive mathematical models for cement kilns due to their high energy
requirements during clinker transformations and environmental issues coming up in recent
times. Apart from aiding in developing a better understanding of complex processes in
cement kilns, mathematical models can also be useful in exploring different means to reduce
energy consumption in the kilns.
The energy consumption in cement kilns is attributed to several processes occurring within
the kiln, namely for raising the temperature of the solids to the reaction temperature, for
endothermic clinkerization reactions, for melting of solids. Apart from these, energy is lost by
the kiln to environment and other connected equipment. The energy of hot solids and hot
gases leaving cement kiln is recovered (at least partially) in clinker cooler and calciner/ pre
heaters. The energy lost to the surroundings from kiln surface (convection and radiation) is
usually not recovered and is of the order of about 15-20 % of total energy input (Engin and
Ari, 2005). Therefore, every attempt should be made to reduce the losses from kiln shell
(without jeopardizing the product quality) per unit weight of the product to enhance kiln
energy performance.
There can be several ways for reducing net losses from kiln shell per unit weight of product.
Increasing the throughput of the cement kiln might be one of them. However, in order to
ensure the same quality of product, it is essential to understand behavior of kiln at different
throughputs. It is desirable to have a computational tool, which allows simulation of kiln
performance over a wide range of design and operating parameter space. One of the obvious
ways of reducing losses from kiln shell can be use of insulation over the kiln shell. However,
such insulation of a kiln shell will lead to a very high temperature of the shell wall (if the kiln
internal temperature was kept the same). This is not acceptable from the mechanical integrity
of the kiln shell. Engin and Ari (2005) have proposed a use of a secondary shell over a rotary
cement kiln to reduce losses from the kiln (see Figure 6 of Engin and Ari (2005)). However,
36
this proposal leads to substantial increase in temperature of kiln wall, which is usually not
acceptable. This suggestion can be made practical by recovering energy from the kiln shell
without reducing the energy flux across the shell. The secondary shell and the interstitial
space can be used as a heat exchanger to recover energy from kiln wall. By selecting
appropriate configuration of a secondary shell and appropriate airflow rate through the
interstitial space, net energy loss to the surrounding can be minimized without altering the
energy flux through kiln wall. The hot air coming out of the annular space could be utilized
somewhere else in the plant (for example drying fly ash or pre-heating feed/air).
In this chapter it is attempted to develop comprehensive mathematical model to simulate

reactions heat transfer and flow simultaneously in a single framework. This model was
further used to realize the promise of using a secondary shell or to explore design and
operating parameter space to minimize energy consumption per unit weight of product. We
have developed a comprehensive model comprising three sub-models viz. model for
simulating variation of bed height in the kiln, model for simulating reactions and heat transfer
in the bed region and model for simulating coal combustion and heat transfer in the freeboard
region for simulating cement kilns. Individual sub-models were validated wherever possible.
Variation of bed height, melting and formation of coating within the kiln were accounted for
the first time simultaneously. Combustion of coal in the freeboard region was modeled in
detail. The model was used to understand influence of various design and operating
parameters on kiln performance. The simulated results were also compared with the available
data from 3 industrial kilns. Several ways of reducing energy consumption in kilns were then
computationally investigated.
3.2 Computational Model and Solution Methodology
In this work, separate models for bed and freeboard region of kiln were developed. These two
models were solved with segregated solution approach and were coupled to each other, along
with a model for bed height via mass and energy communication through common
boundaries. Individual sub models are discussed in the following.
3.2.1 Model for calculating bed height in rotary kilns
Developing accurate models for axial motion of solids in rotary kilns is essential since it
controls the residence time of solid particles and the variation of solid bed height within the
37
kiln. The bed height in rotary kilns depends on volumetric flow rate of solids, φv, kiln tilt (γ,
radian), angle of repose of solids (β, radian), radius of kiln (R, m), rotational speed of kiln (n)
and height of solids (h) (Spurling et al., 2001; Lebas et al., 1995; Kramers and Crookewit,
1951). These parameters are shown in Figure 3.
φv
x h
β ?γ
Figure 3.0: Various parameters influencing bed height in rotating kilns
In Chapter 2, it was discussed that the Kramer’s model (Kramers and Croockewit, 1952) was
shown to predict published experimental data with reasonable accuracy. However, it should
be noted that Kramer’s model does not take into account the effect of counter current gas
flow on axial motion of solids. Fortunately, gas flow does not significantly affect solid hold
up in kilns (Friedman and Marshall, 1949). Therefore, in this work we have used the
Kramer’s model for simulating axial motion of solid particle, which is given by (Kramers and
Croockewit, 1952):
dh ⎡ tan β 3φ v ⎛ 2h h 2 ⎞
−3
2 ⎤
= tan γ ⎢ − ⎜⎜ − 2 ⎟⎟ ⎥ (3.1)
dx ⎢ sin γ 4π nR 3 ⎝ R R ⎠ ⎥
⎣ ⎦
From the above equation height of solids at different axial locations can be calculated. Based
on the height at particular axial position, area of the clinker bed can be calculated as:
1 1
Acl = × R 2 × Γ − × L gcl × ( R − h ) (3.2)
2 2
38
In the above equation Γ is the angle in radians made by the solids at the kiln center and Lgcl is
the cord length of solids in the transverse section. Correspondingly area of freeboard region
was calculated as:
Ag = Ak − Acl (3.3)
3.2.2 Bed model

It is important to consider solids mixing while developing mathematical models for
simulating the cement kilns. Recently, Sherritt et al. (2003) have reviewed axial mixing of
solids in rotary kilns and have proposed a correlation to estimate axial mixing. The
correlation is valid for wide range of solid materials and kiln dimensions. This study
indicated that the values of Peclet number for industrial rotary cement kilns were greater 104.
For such a high values of Peclet number, axial mixing could be neglected as discussed in
previous chapter. Therefore, the mass conservation equation for species i was written
assuming a plug flow of solids as:
d ( AclVcl ρclYi)
= Ri Acl (3.4)
dx
where Yi is the species mass fraction and Ri is the rate of reaction of individual species. Acl is
the area of the clinker bed normal to the kiln axis, Vcl is the velocity of the solid bed, ρcl is
bulk density of the solids. By summing the conservation equation over all the species, the
overall mass conservation equation can be written as:
d ( Acl Vcl ρ cl )
= − RCO 2 Acl (3.5)
dx
RCO2 is the rate of formation of CO2. CO2 formed due calcination reaction in the bed region
leaves the bed. Hence the solids flow rate in the kiln decreases along the kiln length.
Calculation of RCO2 is discussed later in this section. The area of the solids bed Acl was
calculated by Equation (3.2). The variables Γ and Lgcl are functions of bed height in the kiln
and can be calculated once the bed height is known. The bed height in the kiln varies along
the kiln length. The variation of bed height within the kiln was modeled using Equation (3.1).
The chemical reactions occurring during clinker formation were discussed in Chapter 2
(Equation 2.3 to 2.7). The rate expressions for these reactions were calculated as:
39
NC NR Z ij − Ej NC
Ri = ∑ ∑ )∏ Ck
o( j,k )
k0 j exp( (3.6)
i =1 j =1 Mj RTcl k =1
Here, Zij are the stoichiometric coefficients, Mj is the molecular weight of the base
component, k0j and Ej are the Arrehenius parameters of the reactions. Tcl is the bed
temperature, Ck is concentration of the kth reactant and o (j, k) is order of jth reaction with
respect to kth component. We have assumed the reactions to be of first order with respect to
each reactant. NR is the number of reactions, which vary from 1 to 5, and NC is the number
of components, which vary from 1 to 10 in the present case. The solids (particles) were
assumed to be pseudo homogeneous fluids as explained in the previous section.
Selecting appropriate kinetic parameters for above reactions is important task. Critical review
of published literature pointed out that there are significant differences in the kinetics
reported even for relatively simple reaction like limestone calcination. The reported
activation energy for calcination varies from 180-210 kJ/ mol (Irfan and Dogu, 2001) to ~
1500 kJ/ mol (Hills, 1968). Similar level of disagreement was observed even in reported
experimental data. For example, Watkinson and Brimacombe (1982) showed the initiation
temperature of calcination reaction as 1100 K while Irfan and Dogu (2001) reported it as 850
K. Such disagreement in the reported data might be due to differences in the environment of
calcination decomposition (air, N2 or CO2 containing flue gas). Several attempts have been
made to model calcination kinetics. Limestone calcination was shown to be a shrinking core
process (Khraisha et al., 1992). Irfan and Dogu (2001) have carried out TGA experiments for
wide range of limestone species. Several models like shrinking core model with mass transfer
control, shrinking core model with diffusion control, shrinking core model with surface
reaction controlling were compared with the measured experimental data. Shrinking core
model with surface reaction controlling was reported to be the best compared to other models
tested by them. Shrinking core model has also been used by various other investigators to
model limestone calcination (Ingraham and Marier, 1963; Satterfield and Feakes, 1959).
Hence we have used shrinking core model for limestone calcination in the present work. The
reduction in radius of lime stone particle was related as:
−E
dR P M k
= − W s 0 e RT (3.7)
dt ρb
40
Where ρb is the bulk density of the particle, Rp is the particle diameter, Mw is the molecular
weight of the particle, ks0 is the pre exponential factor in the Arrehenius equation and E is the
activation energy.
Equation describing rate of reduction of radius of a single particle was re-written as an

effective rate expression for pseudo homogeneous kinetics as:
−E
dC c 2/3
− = k app C C e RT (3.8a)
dt
where
1 2/3
3ε 3 k ⎛ MW ⎞
k app = S S0 ⎜⎜ ⎟⎟ (3.8b)
R0 ⎝ ρB ⎠
It can be seen that the apparent rate constant depends on the initial particle diameter and the
bulk density of solids. As specified earlier, a wide discrepancy is observed in the proposed
rates for calcination reaction. The large scatter in calcination kinetics was also reported by
Stanmore and Gilot (2005) in their recent review on limestone calcination. It was also
reported that the proposed models fit adequately their own experimental data. However, when
they are applied to other results most of them fail. In absence of any reliable information on
calcination kinetics, in the present work, we have compared models proposed by various
investigators (see Table 3.1) with available experimental data and chosen a set of kinetics that
best fits the data. The data reported in experiments of Watkinson and Brimacombe (1982)
were close to actual kiln operations (bed temperature ~ 1000 to 1300 K) and was used to
select appropriate calcination kinetics. The kinetic models listed in Table 3.1 were
incorporated in the present model to simulate experiments of Watkinson and Brimacombe
(1982). The simulated profiles of percentage calcination in calciner as a function of bed
temperature (along the kiln length) for these different kinetics are shown in Figure 3.1. It can
be seen from Figure 3.1 that the kinetics proposed by Rao et al. (1989) gives the best fit in
the considered group and was therefore used in the present study.
41
Table 3.1: Kinetics proposed by various investigators for calcination reaction
Sr. No Activation Energy Pre-exponential factor Reference
2
(kJ/mol) (kmol/m s)
1 186.9 1.020 × 108 Irfan and Dogu (2001)
2 193.6 9.670 × 107 Irfan and Dogu (2001)
3 160.0 3.650 × 107 Irfan and Dogu (2001)
4 205.8 1.070 × 108 Irfan and Dogu (2001)
5 172.2 6.050 × 107 Irfan and Dogu (2001)
6 156.2 4.190 × 107 Irfan and Dogu (2001)
7 159.1 4.090 × 107 Irfan and Dogu (2001)
8 213.3 1.430 × 108 Irfan and Dogu (2001)
9 166.0 6.700 × 103 Garcia-Labiano et al.(2002)
10 205.0 6.078 × 104 Borgwardt (1985)
11 169.0 1.185 × 103 Ingraham and Marier (1963)
12 154.1 2.230 × 102 Lee et al. (1993)
13 164.7 5.290 × 102 Lee et al. (1993)
14 185.0 1.180 × 103 Rao et al. (1989)
15 131.0 2.540 × 10-1 Garcia-Labiano et al. (2002)
16 205.0 2.470 × 102 Khinast et al. (1996)
42
100
D ata #
90 (1-8) Volume fraction solid in
1* freeboard = 1%
kG = 0.085 W/m K
80 2* (9) εG = 0.1
3* kSolid = 0.69 W/m K
70
4* εSolid = 0.9
kRef = 0.4 W/m K
5*
60
% Calcination
9*
50 10*
(10-13)
11*
40 12*
(14)
13*
30
14*
15*
(15-16)
20
16*
10
0
100 1000 10000
B e d T e m p e r a tu r e , K
Figure 3.1: Comparison of different reaction rates on experimental data of Watkinson

and Brimacombe, (1982); * Numbers in figure correspond to sequence in Table 3.1;
#
Only 5 curves are shown to avoid clutter, rest fall in the same range.
This set of experimental data was chosen for fitting exercise since the experimental
conditions were close to actual kiln operations. Out of total 18 investigations, kinetic
parameters proposed by Rao et al. (1989) gave the best fit and therefore used to model
calcination reaction (See Table 3.1). The experimental conditions of Rao et al. (1989) were
also close to operating conditions in kiln (Temperature in range of 953 K- 1148 K with an
average grain size of ~ 10 µm). Since particle level models are not yet well established for the
rest of the reactions, kinetics of these reactions were used as reported by Mastorakos et al.
(1999) and are specified in Table 3.2.
43
Table 3.2: Reaction, kinetics and heat of reaction
Reaction k0 E ∆H*
(kJ /mol) (kJ/mol)
Bed
1. CaCO3 = CaO + CO2 1.18 × 10 3 (kmol/m2-s) 185 179.4
2. 2CaO + SiO2 = C2S 1.0 × 10 7 (m3/kg-s) 240 -127.6
3. C2S + CaO = C3S 1.0 × 10 9 (m3/kg-s) 420 16.0
4. 3CaO + Al2O3 = C3A 1.0 × 10 8 (m3/kg-s) 310 21.8
5. 4CaO + Al2O3 + Fe2O4 = C4AF 1.0 × 10 8 (m6/kg2-s) 330 -41.3
Freeboard
1 CH4 + 2O2 = CO2 + 2H2O 1.6 × 1010 (m3/kg-s) + 108+ 3125+
2 C + O2 = CO2 1.225 × 103 (m/s) + 99.77+ -
+
Li et al. (2003); * Drujic et al. (2005)
The chemical reactions occurring in the bed are driven by energy supplied by the free board
and kiln walls. Although heat transfer occurs by all the three mechanisms, radiation is a
dominant mode of heat transfer (Mastorakos et al., 1999; Karki et al., 2000) There was a
considerable work on heat transfer in rotary kilns (Gorog et al., 1981; Gorog et al., 1983;
Barr et al., 1989). Gorog et al. (1983) have reported comparison of results with and without
considering contributions of radiative transfer over axial distances. The relative contribution
of radiative transfer over axial distances depends on axial flame temperature gradient
(∆T/∆Z), relative flame diameter (diameter of flame/diameter of kiln) and wall refractivity.
Considering the typical values of these parameters occurring in rotary cement kilns
(temperature gradient of 100 K/m, relative diameter of about 0.5 and wall refractivity in the
range of 0.2-0.5), it can be concluded that the maximum errors caused by neglecting radiative
transfer in axial direction is less than 10%. Based on these results and considering the main
44
objective of this work as developing a simple yet adequately accurate model (without
increasing demands on computational resources), we have neglected the contribution of
radiative transfer over axial distances. The energy conservation equation was written as
NC
d
( AclVcl ρ cl C p Tcl + AclVcl ρ cl λm L ) = ( L gcl Q ' ) − ( ∑ Ri H i ) Acl − S CO2 (3.9)
dx i −1
Cp is heat capacity of solids. The first term of the left hand side represents convective transfer
of energy. The second term of left hand side represents the energy required for melting of
solids. The second term in the RHS of the above equation represented energy required/
liberated during chemical reactions (Hi is the heat of formation of species i). Q´ is the heat
flux received by the bed from the free board and kiln walls and Lgcl is the chord length of
solids exposed to the freeboard region. SCO2 is the volumetric energy sink due to CO2 loss
from the bed. The first term in the RHS of Equation (3.9) represents the heat received by the
solid bed from the freeboard gas and the hot walls. This term was evaluated as explained in
the following.
Various modes by which heat transfer takes place in rotary kiln operation are shown in Figure
3.2. In Figure 3.2, region 1 represents the coating layer, region 2 represents the refractory
layer and region 3 represents the kiln shell. Region 4 represents the annulus or interstitial gap
between the two shells, 5 presents the secondary shell and 6 represents the insulation layer. In
absence of secondary shell layers 4, 5 and 6 would not be there. Energy is transferred from
freeboard to bed via radiation and convection from freeboard gas. Also there will be
exchange of energy between bed and kiln inner walls via radiation and conduction. Hence, to
calculate the amount of energy transferred to the bed, temperature of inner wall and freeboard
gas are required. To calculate the heat losses due to radiation and convection from the kiln,
the outer shell temperature is required. To calculate the temperature of inner walls, refractory,
shell and secondary shell steady state heat balance across the kiln walls need to be solved.
The steady state equations for heat balance across kiln walls were written as:
QRGW + QCGW − QRWB - QCWB = QCOAT (3.10)
QCOAT = QREF (3.11)
QREF = QSHELL (3.12)
45
QSHELL = QRAD + QCONV − SHELL + QRAD − SHELL (3.13)
QRAD = QCONV − SEC − SHELL + QCOND − SEC − SHELL + QRAD − SEC − SHELL (3.14)
QCOND − SEC − SHELL = QINSU (3.15)
QINSU = QLOSSCONV + QLOSSRAD (3.16)
Q RGW + Q CGW Q LOSSCONV

Q RWB
1 2 3 4 56
Q RGB + Q CGB
Q CWB Q LOSSRAD
Bed
Figure 3.2: Heat transfer in transverse section of rotary kiln
In the above equations, QRWB and QCWB are heat transfer rates due to radiation and
conduction respectively between the kiln internal wall and the bed. QRGW and QCGW are heat
transfer rates due to radiation and convection between the freeboard and kiln internal wall.
QCOND-COAT, QCOND-REF, QSHELL, QCOND-SEC-SHELL and QCOND-INSU are the conductive heat
transfer rates in coating, refractory, kiln shell, secondary shell and insulation layers
respectively. QRAD is the radiative heat transfer from kiln shell to secondary shell. QCONV-
SHELL and QCONV-SEC-SHELL are the heat transfer rates between hot shell walls and the gaseous
stream in the annulus. QLOSSCONV and QLOSSRAD are the losses from outer shell due to
convection and radiation respectively. The calculation of individual heat transfer terms in
above equations is given below
46
Heat transfer by radiation between gas phase & bed and gas phase & the kiln internal walls
was evaluated by the equations developed by Hottel and Sarofim, (1967) valid for ε ≥ 0.8:
⎛ ε T 4 − α G TKG ⎞
QRGK = σ ARKG (ε K + 1)⎜⎜ G G ⎟⎟ K = B, W (3.17)
⎝ 2 ⎠
QRGK is the radiative heat transfer, σ is the Stefan- Boltzmann constant, A is the area of heat
transfer. ε and α are the emissivity and absorptivity of the freeboard gas respectively. Tg is
the temperature of the freeboard gas. Subscript B denotes solid bed and W denotes walls.
Convective heat transfer between gas phase & bed and gas phase & the kiln internal walls
was evaluated as:
QCGK = hCGK ACKG (TG - TK ) K = B, W (3.18)
Tscheng and Watkinson (1979) have studied the effects of convective heat transfer of kiln
operating parameters like kiln rotation and kiln degree of fill in pilot plant scale. Based on
their experimental data obtained for sand and limestone, they have proposed correlations for
convective heat transfer coefficients for sand and limestone as:
kg
hCGS = 0.46 Re 0D.535 Re ω0.104 η −0.341 (3.19)
De
kg
hCGW = 1.54 Re 0D.575 Re ω−0.292 (3.20)
De
where kg is the gas thermal conductivity and De is the hydraulic diameter of the kiln. ReD and
Reω are the axial and angular Reynolds number calculated as:
ρg × ug × De ρg × ω × De2
ReD = Reω =
µg µg
(3.21)
0.5D × ( 2π − β + sin β)
De =
β β
(π − + sin )
2 2
It is important to note that the Reynolds numbers for commercial kilns are significantly
higher as compared to experimental conditions of Tscheng and Watkinson (1979). However,
there are no other systematic experimental studies reported to predict convective heat transfer
47
coefficients in industrial rotary kilns. Equations (3.19) and (3.20) were used to model
convective heat transfer in commercial kilns and have reproduced temperature profiles for
gas and solids in these kilns reasonably well (Martins et al., 2000; Martins et al., 2001).
Therefore, Equations (3.19) and (3.20) were used to predict convective heat transfer
coefficients. Fortunately, radiation dominates the overall heat transfer process and therefore
possible errors associated with Equation (3.19) and (3.20) are not expected to change the
simulation results significantly.
The radiative heat transfer between kiln internal walls and bed is given by:
Q RWB = σ × A RWB × ε S × ε W × Ω × (TW4 − TS4 ) (3.22)
Ω is the form factor for radiation which was calculated as:

L gcl
Ω= (3.23)
(2π − β ) R
The heat transfer by conduction between the bed and the kiln internal walls was given by:
QCWS = hCWS ACWS (TW - TS ) (3.24)
hCWS is the heat transfer coefficient which was evaluated using empirical correlation by
Tscheng and Watkinson (1979) as:
0. 3
k ⎛ ωR 2 Γ ⎞
hCGW = 11.6 b ⎜⎜ ⎟⎟ (3.25)
Acws ⎝ α b ⎠
In the above expression kb is the thermal conductivity of bed, Γ is the angle of fill of the kiln,
ω is the rotational speed (rad/s) and αb is the bed thermal diffusivity. Dhanjal et al. (2004)
also have presented their work on wall to bed heat transfer wherein experimental results were
presented by heating sand of different particle size distribution in a batch rotary kiln to bed
temperatures up to 775 0C. However, Martins et al. (2001; 2000) have used and shown that
correlations (from Equation 3.17 to Equation 3.25) work reasonably well for industrial as
well as lab scale kilns and were used in this work.
Conductivity and emissivity of the freeboard gases depend on its composition and dustiness
(volume fraction of solids entrained in the freeboard gases). The influence of carbon dioxide
48
and water vapor on emissivity was estimated from the data reported by Gorog et al. (1983)
The thermal conductivity of the freeboard gas was assumed to be same as that of air at
prevailing temperature and was calculated by the expression:
2 3
k g = - 7.494 × 10-3 + 1.709 × 10-4 Tg - 2.377 × 10-7 Tg + 2.202 × 10-10 Tg
4 5
(3.26)
- 9.463 × 10-14 Tg + 1.581 × 10-17 Tg
The thermal properties of solids were available from literature and are specified in Table 3.4
Influence of entrained solids on effective emissivity and conductivity of gas was estimated
as:
ψ eff = ε solidψ solid + (1 − ε solid )ψ gas (3.27)
where ψ is the property i.e. thermal conductivity or emissivity, εsolid is the solid volume
fraction or dustiness of the gas. Adequate information on solid entrainment or solid volume
fraction in rotating cylinders is however not readily available. Few experimental studies on
solid entrainment on rotary kilns (Friedman and Marshall, 1949; Li, 1974) suggest that the
volume fraction of solids in freeboard region of kiln is very small (<1 %). Therefore, for all
the subsequent simulations, the value of solid volume fraction entrained in the free board
region was set to 1 %. Fortunately, the predicted results (C3S mass fraction and temperature
profiles of bed and freeboard) were not found to be sensitive to the value of solid hold-up in
the freeboard region up to 5 %.
QRAD is the radiative heat transfer between shell and secondary shell. Since the secondary
shell is considered to be of material of low emissivity, radiation from secondary shell to kiln
shell is neglected (Engin and Ari, 2005). The radiative heat transfer between of shell and
secondary shell was calculated as (Engin and Ari, 2005):
A σ (TSHELL − TSEC − SHELL )

4 4
QRAD = SHELL (3.28)
1 1 − ε 2 ⎛ RSHELL ⎞
+ ⎜ ⎟
ε1 ε 2 ⎜⎝ RSEC −SHELL ⎟⎠
QRAD-SHELL and QRAD-SEC-SHELL are the radiative heat transfer rates between hot shell walls and
air in the shell annular gap and were calculated as:
49
QRAD − SHELL = σ × ASHELL × ε SHELL × ( TSHELL
4
− TAIR
4
) (3.29)
QRAD − SEC − SHELL = σ × ASEC - SHELL × ε SEC − SHELL × ( TSEC

4
− SHELL − TAIR )
4
(3.30)
QCONV-SHELL and QCONV-SEC-SHELL are the convective heat transfer rates between the hot shell
walls and air passing through interstitial gap. The convective heat transfer rates are given by
QCONV-SHELL = UCONV × ASHELL × (TSHELL - TAIR) (3.30a)
QCONV-SEC-SHELL = UCONV × ASEC-SHELL × (TSEC-SHELL - TAIR) (3.30b)
Depending on mass of air through annulus, the heat transfer coefficient for convection for a
turbulent flow (Re> 2300) was calculated as: (Quirrenbach, 1960)
d o 45
U CONV = 0.021 (Re) 0.8 (Pr) 0.33 ( ) (3.30c)
di
In absence of gaseous stream through the annular space between the two cylinders the terms
QCONV-SHELL and QCONV-SEC-SHELL would be zero.
The heat losses by convection and radiation in presence of secondary shell with insulation
were calculated as:
QCONVLOSS = HCONV × ASHELL × (TINSU – T0) (3.31a)
Q RADLOSS = σ × ASHELL × ε SHELL × ( TINSU

4
− T04 ) (3.31b)
In absence of secondary shell the heat losses were calculated as:
QCONVLOSS = HCONV × ASHELL × (TSHELL – T0) (3.32a)
Q RADLOSS = σ × ASHELL × ε SHELL × ( TSHELL

4
− T04 ) (3.32b)
Above set of non-linear equations with appropriate boundary conditions were solved to get
the temperatures of kiln walls (Equations (3.10) to (3.12) for the case without a secondary
shell and Equations (3.10) to (3.16) for a case with a secondary shell). Knowing the
temperature of kiln inner wall and freeboard gas, the heat flux received by the bed (Q' in
Equation 9) was calculated as:
50
QCWB QRWB QRGB QCGB
Q' = + + + (3.33)
ACWB ACWB ACWB ACWB
Here QRGB and QCGB are heat transfer due to radiation and convection between freeboard gas
and bed respectively.
Melting of solids in the kilns was assumed to be proportional to the bed temperature in the
kiln. Melting was modeled by using the following relationship between mass fraction of
liquid and bed temperature:
⎡ T − TS ⎤
m L = max ⎢0 , cl ⎥ (3.34)
⎣ TL − TS ⎦
where TL and TS are liquidus (temperature at which all mass is liquid) and solidus
(temperature at which first drop of liquid forms) temperatures respectively. λ is the latent heat
of melting (λ = 416 kJ/kg) (Mastorakos et al., 1999). The formation of melt phase was
assumed to depend only on local temperature (it was assumed to occur very fast compared to
other relevant time scales). The values of TL and TS depend on composition of solids in the
kiln and may vary within the kiln. As a first approximation, these values were assumed to be
constant and were set to 1560 K and 2200 K respectively following Peray (1986).
As the part of the solids melt in the kiln, the melt forms an inner coating on the refractories.
The coating formation will reduce the effective diameter of the kiln and will reduce heat
losses. The change in an effective diameter of the kiln will change effective contact area
between the solid bed and the free board and may affect profile of bed height within the kiln.
The coating formation is an extremely complex process, which depends on flame
temperature, chemical composition of kiln feed, temperature of coating and temperature of
bed. The mechanism of coating formation and its interaction with design and operating
parameters are not well understood. In practice, usually coating thickness is measured under
cold conditions. Some of the data on maximum coating thickness available from industrial
sources (listed in Table 3.2) and the values reported by Mastorakos et al. (1999) and
Karki et al. (2000) indicate that the maximum coating thickness varies within the range of
0.03DI < CT,max < 0.05DI, where DI is the internal diameter of the refractories. The observed
differences in the predicted results (temperature profiles) with the minimum (0.03DI) and
51
maximum (0.05DI) values of CT,max were within 1%. Therefore, in this work, if the measured
value of maximum coating thickness is available, it was used as an input data. In absence of
such information, the maximum coating thickness was set to 0.04 DI.
As mentioned earlier, coating starts forming when bed temperature exceeds solidus
temperature and attains a maximum thickness after some point. In absence of any better
information, in this work, we have assumed that the coating thickness varies linearly with the
bed temperature up to certain limit and then remains constant thereafter. Thus the coating
thickness (CT) along the kiln length was calculated as:
CT = 0.0 for Tcl < Ts
⎛ T − TS ⎞
CT = ⎜⎜ cl ⎟⎟ CT ,max for Ts < Tcl < Tcoat (3.35)
⎝ Tcoat − TS ⎠
CT = CT, max for Tcl > Tcoat
TS is the bed solidus temperature at which bed starts melting. Tcoat is a temperature at which
coating is assumed to attain its maximum thickness, CT,max. The value of Tcoat was set to 1600
K. It was confirmed that changing the value of Tcoat by 5 % had an insignificant (within 1%)
effect on temperature profiles of bed and freeboard gas.
3.2.3 Freeboard model
Coal particles enter the freeboard region along with secondary air from the burner end. The
pulverized coal particles enter the kiln via high velocity jets (about 50 m/s) where as
superficial velocity based on kiln cross section of secondary air is within the range of 10-15
m/s. Considering the large difference in the actual velocity of coal particles and superficial
gas velocity in the kiln and inertia of coal particles, it is inappropriate to assume coal particle
velocity same as that of superficial velocity near the burner. Such an assumption would have
significant implications for simulation of devlotalization and combustion of coal particles
(Gang et al., 2000, Megalos et al., 2001). Away from the burner, the velocity of coal particle
will eventually be more or less as the gas velocity. It is necessary to make approximations for
the decaying profile of particle velocity within the one-dimensional framework of the model.
In this work, the axial velocity decay of coal particles was calculated using a semi empirical
equation proposed as Megalos et al. (2001)
u c = u c , 0 e −δ z (3.36)
52
where uc,0 is the coal velocity at the nozzle entrance. Equation (3.36) was used un till the
velocity of coal particles was greater than the superficial gas velocity. Once particle velocity
reached the superficial gas velocity it was assumed to move with the superficial gas velocity
in the kiln. It is important to note that Equation (3.36) is an empirical system specific
equation. The parameter δ appearing in Equation (3.36) is a characteristic constant for a
specific burner/ kiln configuration. In absence of any guidelines for determination of values
of δ for kilns, in this work we had estimated the values of δ from CFD simulations of the free
board region developed in Chapter 6. The particle velocities predicted by the CFD model
were fitted to Equation (3.36) to obtain the value of δ. A value of 0.06 gave a reasonably
good fit. Fortunately, the predicted results are not very sensitive to the value of δ (± 20 %
variation in the value of δ resulted in less than 0.1 % change in the freeboard gas temperature
profile along the kiln length).
The present computational model assumes coal particles to be comprised of char, volatiles
and ash. Coal combustion was modeled using a shrinking core model with constant
devolatilization, finite rate chemistry and char combustion. While developing the coal
combustion model, it was assumed that, devolatilization proceeds with shrinking core of char,
which because of relatively high porosity does not offer any resistance to gases leaving the
reaction zone (Chern and Hayhurst, 2004). The coal particle first gets heated by absorbing
energy from the gas. The energy is received by convection and radiation. Due to heat up the
coal particle starts devolatilization. The present model assumes a constant rate
devolatilization for coal combustion (Pillai, 1981). Coal devolatilization rate is given as:
dFc ,vol − A0 FC 0 mv0

= (3.37)
dz uc
Here Fc,vol is volatilization rate of coal particle (kg/s), FC0 is the initial coal flow rate and mv0
is the initial mass fraction of the volatiles in the coal. uc is the velocity of the coal particle. A0
(s-1) is the devolatilization constant recommended as 12.0 for coal combustion. The coal
particle density was assumed to decrease in proportion to the volatiles released in the gas
phase (Heidenreich et al., 1999). On devolatilization, in the gas phase, volatiles will react
with available oxygen in the freeboard region to form CO2 and H2O (Guo et al., 2003;
Martins et al., 2001). Gas phase combustion was modeled as reported by Guo et al. (2003) for
53
pulverized coal combustor. A global single step reaction of volatile fuel was assumed to take
place as:
CH 4 + 2O 2 → CO2 + 2 H 2 O (3.38)
The kinetics and heat of reaction for above gas phase reaction are given in Table 3.1. It
should be noted that the characteristic reaction time scale at temperatures above 1500 K is
less than 1 ms. Therefore, at such high temperatures, the reaction will be limited by turbulent
mixing rather than reaction kinetics. For turbulent mixing limited combustion, the most
commonly used model is the eddy breakup (EBU) turbulent combustion model of Magnussen
and Hjertager (1976). This model compares the kinetic rate and the turbulence mixing rate
and selects the lower rate for further calculations as:
Rcombg = min (RS, EBU, RS, Arr) where
k y ox
RS , EBU = C R ρ min( y F , ),
ε b (3.39)
E
RS , Arr = Bs ρ y F y OX
2
exp(− s )
RT
yF is the mass fraction of the fuel and yOX is the mass fraction of oxygen. The values for
Arrhenius kinetic constants BS and ES are specified in Table 3.1. CR is a parameter of the eddy
breakup (EBU) turbulent combustion model, which is usually set to 4. It can be seen that the
EBU model uses ratio of turbulent kinetic energy and energy dissipation rate (k/ε) as
characteristic time scale of turbulent mixing. In the present one-dimensional reaction
engineering framework for modeling cement kilns, turbulent flow in the free board region is
not being simulated. Therefore it is necessary to estimate the variation of (k/ε) along the kiln
length. CFD simulations developed in course of this work were used to understand variation
of (k/ε). The results indicated the possibility of approximating the variation by a linear
profile. A simplified analysis of one-dimensional version of modeled equations of k and ε
(neglecting accumulation, turbulent dispersion and approximating values of model
parameters) lends some support to the linear approximation. Based on this, the variation of
(k/ε) was approximated as:
(k/ε) = a (z/L) [a = 1 s] (3.40)
where z is distance from the burner end and L is kiln length. Equation 3.40 represents system
specific empirical one-dimensional version of modeled equations for k (turbulent kinetic
energy) and ε (dissipation rate) along the kiln length. It can be seen that near the burner end,
54
time scale of turbulent mixing might be smaller than the reaction time scale (due to lower
temperature and higher turbulence at the burner end). However, in the major portion of the
kiln, the effective rate will be controlled by the rate of turbulent mixing.
Single step combustion reaction was assumed for char combustion in the present model
(Spang, 1972):
C + O2 → CO2 (3.41)
For the temperature conditions and range of coal particle size (~ 50 µm) found in cement
kilns, the reaction rate was written as (Smith, 1971; Young and Smith, 1989),
M Coke N P AP k sCO2
Rcomb = (3.42)
Ag uc
Where Rcomb is the combustion rate per unit volume of the gas (kg/m3 s), Mcoke is the
molecular weight of coke, AP (m2) is the surface area of each particle. ks (m/s) is the rate
constant for a first order reaction per unit surface area of the un-reacted core given in Table
3.1, uc (m/s) is the particle velocity and Ag (m2) area of freeboard region. CO2 is the
concentration of oxygen in air. From the above equations the variation of radius along the
axis of the freeboard can easily shown to be derived as:
dr 1 Co
= ks . 2 (3.43)
dz u C CC
Where z is the distance from gas inlet along the axis and CC is the concentration of carbon in
coke. Now we present the mass and energy balance equations for discrete (coal) and
continuous (gas) phase for the freeboard region.
Mass balance for discrete phase
The mass balance for discrete phase is given by
dFC dFc ,vol

= −υ coke Rcomb Ag − (3.44)
dz dz
Where FC is the mass flow rate of coke and νcoke the stoichiometric coefficient of coke in the
combustion reaction and Ag is the area of the freeboard region. The first term in the RHS of
above equation represents the loss in discrete phase mass due to char reaction and the second
term in the RHS of above equation represents the loss in discrete phase mass due to
devolatilization of coal particles.
55
Energy balance for discrete phase
The energy balance for discrete phase is given as:
d ( FC .C pc .TC ) N P AP hc N σε A
= ( TG − TC ) + P S P ( TG 4 − TC 4 ) + f h ∆H comb Rcomb Ag
dz uc uc (3.45)
In the above equation CPC is the specific heat of coal and TC is the coal temperature. Coal was
assumed to enter the kiln at 300 K. TG is the gas temperature. ∆Hcomb is the heat released due
to coal combustion and fh is the fraction of energy released due to coal combustion, which is
absorbed by the particle. This factor was set to 0.3 following the recommendation of Boyd
and Kent (1986). The first and second terms in RHS of above equations are the energy
absorbed by the coal particle due to convection and radiation from the gas phase during the
heat up. The heat transfer coefficient, hc, is evaluated using the correlation of Ranz and
Marshall (1952) as:
hc d p
= 2 + 0.6(Re) 0.5 (Pr) 0.33 (3.46)
kg
Mass balance for continuous phase
The overall mass balance equation is given by:
dFG dF
= υi Ag Rcomb + c , vol + S G (3.47)
dz dz
Here FG represents the mass flow of gas (kg/s). The first term in the RHS is the mass
transferred to the gas phase due to combustion of coal. The second term in RHS contains the
source terms for the gas phase due to devolatilization of coal particle. The final term in the
above equation is mass of CO2 coming to freeboard region due to calcination reaction in the
bed region.
Mass balance of individual species can be written as
dFGi υ i Ag M Gi .Rcomb dFc ,i

= + +S i (3.48)
dz M coke dz
Where FGi is the mass flow of species i in the gas phase with a stoichiometric coefficient νi,
molecular weight MGi and Si is the source term. Si accounts for species mass transfer to the
freeboard region due to coal devolatilization, char combustion and CO2 formation in the bed
region due to calcination reaction.
56
Energy balance for the continuous phase
An energy balance for the gas phase can be written as.
d ( FG .CPG .TG )
= −∑ Q + ∆H comb .Rcomb . Ag + ∆H combg .Rcombg . Ag + S H
''
(3.49)
dz
where FG is the gas mass flow rate, CPG is the specific heat of gas, TG is the temperature of
the gas phase, ∆Hcomb is the heat of reaction of coke combustion per kg of coke burnt,
∆Hcombg is the heat of reaction of gas phase reaction per kg of volatiles burnt, Ag is the
freeboard area in the transverse section. SH,i is the heat from source/sink to freeboard region.
Freeboard region will receive heat from CO2 due to calcination in the bed region and give
heat to coal particle due to convection and radiation. ∑Q´´ (W/m) is the net heat exchange
between gas and kiln walls/bed. The heat transfer rate was obtained by solving heat balance
across kiln walls in the transverse section as are previously explained in bed model.
Energy balance for convective air in annular gap

Along with the equation for continuous (gas) and dispersed phase (coal) convective air
temperature (air in annular gap of two shells) was calculated as:
d ( M AIR C P ,AIR T AIR )
= U CONV ASHELL ( TSHELL − T AIR ) + σε SHELL ASHELL ( TSHELL
4
− T AIR
4
)
dz
+ U CONV ASEC − SHELL ( TSEC − SHELL − T AIR ) + σε SEC − SHELL ASHELL ( TSEC
4
− SHELL − T AIR )
4
(3.50)
In the above equation MAIR (kg/s) is the mass of air flowing in the annular gap, CP, AIR is the
specific heat of the air. TSHELL and TSEC-SHELL are the temperatures of kiln shell and secondary
shell obtained by solving the heat balance in transverse section of the kiln. ASHELL and ASEC-
SHELL are corresponding areas of heat transfer. UCONV is the convective heat transfer
coefficient calculated by Equation (3.30c). The above set of equations for height variation,
bed and freeboard are solved as described in the next section.
3.2.4 Solution methodology
In this work, all the three sub models viz. height variation model, bed model and freeboard
model were solved separately. The solution procedure employed is shown in Figure 3.3.
Kramer’s equation (Equation 3.1) was solved initially to obtain the profile of bed height
along the kiln length. Using the solution of height variation model, the freeboard model was
solved by assuming a guessed temperature (linear) profile for the solid bed. The solution of
the freeboard model (freeboard temperature) was used to solve the bed model to obtain the
57
new bed temperature profile. This bed temperature was used for solving the freeboard model.
Thus, bed temperature profile was used as a boundary condition to solve the freeboard model.
Similarly, freeboard temperature profile was used as a boundary condition in the bed model.
The other boundary condition for freeboard and bed model was convective and radiative
losses from kiln shell [in absence of secondary shell, Equation (3.32)]. For configuration with
a secondary shell the boundary condition was modified to convective and radiative heat
losses from insulation [Equation (3.31)]. The initial conditions to the freeboard model were
mass and temperature of air exiting from the cooler. The air entering the annular space
between the kiln and the secondary shell was assumed to be at 300 K. The mass and
temperature of solids exiting from the calciner were used to specify the inlet conditions for
the bed model. The modified Gear’s method implemented in ODEPACK was used to solve
the bed and freeboard model equations. The simultaneous solution of non-linear equations
representing the energy balance across the cross section of the kiln (Equation 10 to Equation
16) was obtained using the Newton-Raphson method. This procedure was carried till the bed
and freeboard temperature profiles along the kiln length are within ± 0.1 % on any further
iteration. Typically 15-20 iterations were required for convergence.
Read input data
Height variation using Kramer’s model
Initiate bed temperature

1
Solve freeboard
Get Gas temperature
Solve bed
Get bed temperature
Compare with 1
Not Converged Converged

End
Figure 3.3: Solution Methodology
3.3 Results and Discussion
The computational model described in the previous section was initially used to understand
the behavior of rotary cement kilns. Initially the bed model was validated with available
58
experimental data in published literature. Thereafter, the model was then used to simulate
performance of 3 industrial kilns. After that the model was used to understand influence of
key design and operating parameters on net energy consumption in kilns. In these
simulations, secondary shell around kiln was not considered. The model was thereafter used
to examine possibility of using a secondary shell to reduce energy consumption.
3.3.1 Validation of bed model
Watkinson and Brimacombe (1982) have carried out calcination experiments in a laboratory
scale rotary kiln and have reported temperature profiles within the kiln. Availability of the
experimentally measured temperature profile within the kiln makes this data set particularly
useful to evaluate the one-dimensional model developed here. The experiments of Watkinson
and Brimacombe (1982) were simulated by providing geometrical and operating conditions
of their kiln as input data to the model (as specified in Table 3.3). The measured temperature
profile of the free board region was specified as an input instead of solving the freeboard
equations. Once the temperature profile for gas phase (in free board region) is known, the
two-point boundary value problem reduces to a simple initial value problem and only model
equations for bed were solved. The thermal properties of dusty gases i.e. emissivity and
thermal conductivity used in the industrial kiln simulations were obtained at an average gas
temperature of 1300 K. Gas emissivity for dust free gas was set to 0.1 from the charts
reported by Gorog et al. (1981).
Table 3.3: Dimensions of kilns and operating conditions of kilns
Sr. Variable Pilot Kiln
No.
1. Length, m 5.5
2. Inner refractory diameter, m 0.406
3. Outer refractory diameter, m 0.59
4. Outer shell diameter, m 0.61
5. Coating thickness, m -
6. Speed of rotation, RPM 1.5
7. Angle of tilt, degrees 1
8. Solids flow in, kg/sec 0.013
9. Gas temperature at burner end, K 1600
10. Gas temperature at solid entry, K 800
59
Table 3.4: Physical properties used for lime kiln simulations
Sr. Variable Pilot Kiln
No.
1. Bed Density, kg/m3 1680#
2. Bed Heat capacity, KJ/ kg K 0.8$
3. Bed emissivity 0.9
4. Wall emissivity 0.9
5. Bed Thermal conductivity, W/m K 0.69#
6. Refractory thermal conductivity, W/m K 0.4#
7. Gas heat capacity, (0.106 TG + 1173)!
8. Gas Viscosity (0.1672 × 10 −5 TG − 1.058 × 10 −5 )!
*
Perry (1984); #Barr et al. (1989); +Mastorakos et al. (1999); $Martin, (1932)
!
Guo et al. (2003) ;
1800
E = 185 kJ/mol
1600 5
kapp = 3.15509 x 10 (kmol/m3)1/3sec-1
Kgas =0.085 W/m k
1400 egas = 0.1
ksolid = 0.69 W/m k
esolid = 0.9
1200 Kref = 0.4 W/mk
Temperature (K)
1000
800
600 Experimental Gas points

Experimental Bed Temperature
400
Gas Temperature
200 Bed Temperature (1 % solid volume fraction)

Bed Temperature (0.5 % solid volume fraction)
0
0 0.2 0.4 0.6 0.8 1
Normalized Kiln Length (-)
Figure 3.4: Comparison of experimental and simulated results for bed temperature
In this case, there is only one adjustable parameter (since the freeboard temperature profile is
known): volume fraction of solids in the freeboard. The value of solids volume fraction was
60
initially set to 1 %. The kinetic models listed in Table 3.1 were incorporated in the present
model to simulate experiments of Watkinson and Brimacombe (1982).
With kinetic parameters reported by Rao et al. (1989), the sensitivity of the predicted bed
temperature profile was examined for different values of solids volume fraction in freeboard
region. These results are shown in Figure 3.4. It can be seen that the predicted temperature
profiles are insensitive to the exact values of solids volume fractions within the expected
range. Therefore values of solid volume fraction was set to 1 % for rest of the simulations.
The model was thus able to simulate the experiments of Watkinson and Brimacombe (1982)
reasonably well. The model was then used to simulate performance of industrial cement
kilns.
3.3.2 Simulations of rotary cement kilns

Rotary cement kiln operations involve very high temperatures, which restricts sampling of
data inside the kiln during the kiln operations. The computational model can be used to gain
some insight into complex processes occurring within cement kilns. For this purpose, the
model was used to simulate typical industrial cement kilns. Three different industrial kilns
were selected for simulations. Details of these three kilns are specified in Table 3.5. It can be
seen that the selected kilns cover a reasonable range of kiln dimensions, production capacity
and operational parameters like rotational speed, fill ratio and angle of tilt. Appropriate input
data for carrying out simulations of industrial cement kilns are however not readily available.
The practical difficulties in sampling inside the industrial kilns restricts for available data
only at the pre-heater entrance and clinker cooler exit. The temperature and mass fractions at
the kiln inlet are however, required to specify the initial conditions to the model. A direct way
to obtain these initial conditions would be actual measurements. In absence of such data, in
the present work, the conditions at the kiln entrance were calculated by specifying
appropriate degree of calcination occurring in calciner. In order to identify appropriate degree
of calcination at the kiln entrance, the over all energy balance on the calciner and kiln was
solved for different degrees of calcinations. Knowledge of air temperature at the kiln outlet
(at the solids entrance end) then enabled us to select the appropriate value of calcination. The
mass fractions and temperature of solids and air entering the kiln are given in Table 3.7.
These were used as the initial conditions to the model
61
Table 3.5: Dimensions of kilns and operating conditions of kilns
Sr. Variable Industrial Industrial Industrial
No. Kiln 1 Kiln 2 Kiln 3
1. Length, m 50 60 68
2. Inner refractory diameter, m 3.4 3.6 4.0
3. Outer refractory diameter, m 3.8 4 4.4
4. Outer shell diameter, m 3.85 4.084 4.456
5. Coating thickness, m 0.13 0.144 0.16
6. Speed of rotation, RPM 5.5 3.3 3.5
7. Angle of tilt, degrees 2 3.5 2
8. Solids flow in, kg/s 38.88 50.78 48.12
9. Height at solid entry, m 0.46 0.57 0.7
10 Gas Flow rate kg/s 17 18.26 16.72
11. Gas temperature at burner 1373 1373 1373
end, K
12. Coal Fr, kg/s 1.25 1.33 1.46
13. Char 0.58 0.55 0.58
14. Volatiles 0.27 0.35 0.27
15. Ash Content 0.15 0.1 0.15
16. Calorific value, kcal/kg coal 5800 7000 5000
(kJ/kg coal) (2.436×104) (2.94×104) (2.1×104)
17 Coal Particle size, microns 50 50 50
62
Table 3.6: Physical properties used for cement kiln simulations
Sr. Variable Value
No.
1. Bed Density, kg/m3 1046
2. Bed Heat capacity, kJ/ kg K 1.088
3. Bed emissivity 0.9
4. Wall emissivity 0.9
5. Kiln shell emissivity 0.78
6. Secondary shell emissivity 0.35
7. Bed Thermal conductivity, W/m K 0.5
8. Refractory thermal conductivity, W/m K 4.0^
9. Gas heat capacity (J/kg K) (0.106 TG + 1173)
10. Gas Viscosity (kg/m/s) (0.1672 × 10 −5 TG − 1.058 × 10 −5 )
^
http://www.vrwrefractories.com/cement-industry.html
Table 3.7: Model assumptions/predictions and comparison with industrial data

Sr. Industrial Kiln 1 Industrial Kiln 2 Industrial Kiln 3
No (67 % (60 % (71 %
Calcination in Calcination in Calcination in
calciner) calciner) calciner)
Data Model Data Model Data Model
Model inputs at solid entrance
1. Mass fraction CaCO3
- 0.340 - 0.398 - 0.305
2. Mass fraction CaO - 0.396 - 0.335 - 0.418
3. Mass fraction SiO2 - 0.179 - 0.185 - 0.190
4. Mass fraction Al2O3 - 0.0425 - 0.041 - 0.043
5. Mass fraction Fe2O3 - 0.0425 - 0.041 - 0.043
6. Temperature of solids, K - 1123 - 1250 - 1025
63
The simulations of these cement kilns were carried out based on the model discussed in the
previous section. The angle of repose for solids was set to 0.71 radians based on experimental
data reported by Henein et al. (1983) for calcineous materials. The thermal properties of
gases i.e. emissivity and thermal conductivity were obtained at an average gas temperature of
1750 K. Gas emissivity for dust free gas was set to 0.4 at conditions prevailing in the kiln
from the charts reported by Gorog et al. (1983) The convective heat transfer coefficient for
calculating heat losses was specified to 30 W/m2 K based on work of Mastorakos et al.
(1999) Typical profiles of mass fraction and temperature obtained from simulations for
industrial kiln 1 are shown in Figures 3.5a, 3.5b and 3.5c respectively. For all these figures
abscissa of 0 denotes the position at which solids enter the kiln and abscissa of 1 indicates the
solid exit i.e. the burner end. Figure 3.5a shows the mass fractions of solids in the bed along
the kiln length. After the solids enter the kiln, calcination reaction and formation of C2S take
place. This is followed by C3A and C4AF formation. Solids start melting after traveling about
75 % of kiln length. The amount of melt starts increasing beyond the point where bed
temperature crosses the solidus temperature reaches a maximum and then decreases
thereafter. The mass fraction profiles for the freeboard region are shown in Figure 3.5b. It can
be seen from this figure that the volatiles concentration in gas phase increases as one move
away from the burner end due to coal devolatilization. Thereafter gas phase combustion
starts. Hence H2O concentration in freeboard region increases to about half of the kiln length.
Both coal combustion and gas phase combustion gets completed in half a kiln length. CO2
concentration in the freeboard region near the bed entrance increases due to calcination
reaction in the bed region. The predicted temperature profiles of gas, bed walls and coating
are shown in Figure 3.5c. Coating was formed around 25 % of kiln length from the solid exit
end. The temperature of kiln shell decreases abruptly at the point from which coating
formation occurs. This was due to additional resistance to heat transfer due to coating
formation. Such sharp dips in shell temperature after coating formation (~200 0C) have been
reported in literature for commercial cement kilns (Markatos and Koyefetis, 1995).
The predicted results thus, agree qualitatively with the previously reported results. The
predicted results at the outlet of the kiln were compared quantitatively with the available
plant data in Table 3.8. It can be seen that the model was also able to predict the overall
performance of the industrial kilns satisfactorily.
64
0.7
CACO3
CAO
0.6 C2S
C3S
C3A
0.5 C4AF
Melt
Mass fraction
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Normalised kiln length
Figure 3.5a: Typical mass fractions in bed region
65
0.9
0.8
0.7
0.6
O2
Mass Fraction
0.5 CO2
N2
0.4
Volatiles
0.3 H2O
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Figure 3.5b: Typical mass fractions in freeboard region
66
2500
Bed
Coating
2000 Gas
Inner Wall
Inner Shell
Temperature, K
1500 Outer Shell
1000
500
0
0 0.2 0.4 0.6 0.8 1
Figure 3.5c: Temperature profiles for industrial kiln1
67
Table 3.8: Comparison of model equations with industrial data
Sr. Industrial Kiln 1 Industrial Kiln 2 Industrial Kiln 3
No (67 % (60 % (71 %
Calcination in Calcination in Calcination in
calciner) calciner) calciner)
Model predictions and Data Model Data Model Data Model
industrial data at solids exit
7. Mass Fraction C3S 0.483 0.481 0.508 0.487 0.5 0.489
8. Mass Fraction C2S 0.239 0.238 0.257 0.277 0.269 0.263
9. Mass Fraction C3A 0.051 0.052 0.048 0.052 0.042 0.052
10. Mass Fraction C4AF 0.143 0.146 0.152 0.144 0.142 0.147
11. Mass Fraction CaO 0.084 0.083 0.035 0.04 0.047 0.049
12. Temperature of solids, K 1673 1608 1673 1636 1573 1600
3.3.3 Influence of key design and operating parameters

The model was then used to understand influence of key design and operating parameters of
the rotary cement kiln. Influences of kiln rotational speed and kiln tilt on the predicted outlet
composition of C3S are shown in Figures 3.6a and 3.6b respectively. It can be seen that as the
rotational speed increases, the residence time of solids decreases and hence the outlet mass
fraction of C3S decreases. As angle of tilt decreases residence time of solids increases and
therefore the outlet mass fraction of C3S increases. The predictions of model for residence
time of solids in the kiln are consistent with published results (Liu and Specht, 2006; Li et al.,
2002). Predicted residence time of solids for all these cases are shown in these figures. It
should be also noted that the developed model could be used to study the effect of varying
different operating conditions simultaneously, in order to find optimum operation condition.
For example, the effect of varying kiln tilt and rpm simultaneously on the performance of kiln
can be predicted using the developed computational model. The simulated results were used
to quantify performance of the rotary cement kilns.
68
0.6 2500
0.5
2000
C3S mass fraction
Residence time, s
0.4
1500
0.3
1000
0.2
500
0.1
0 0
2 3 4 5 6 7 8 9
Kiln RPM
Figure 3.6a: Influence of rotational speed on kiln performance (kiln tilt = 20)
0.6 2500
0.5
2100
C3S mass fraction
Residence time, s
0.4
1700
0.3
1300
0.2
900
0.1
0 500
0 0.5 1 1.5 2 2.5 3 3.5
Kiln tilt, degrees
Figure 3.6b: Influence of angle of tilt on kiln performance (kiln rpm = 5.5)
69
3.3.4 Energy consumption in rotary cement kilns
The kiln performance was characterized in terms of net energy consumption per unit weight
of product (clinker coming out of the kiln) for the same (within ± 1%) product composition.
The net energy consumption was calculated as (See Figure 3.7):
Net energy consumption = (ECOAL + EG,IN + ES,IN) – (EG,OUT + ES,OUT) (3.51)
ELOSS ECOAL
EG, OUT EG, IN
Cement Kiln
ES, IN ES, OUT
Figure 3.7: Energy consumption in cement kilns
ECOAL is the energy given by coal combustion per unit product. Other terms denote energy
flow rates (subscripts IN or OUT) for the gas and solid streams (subscripts G or S) per unit
weight of product. Influence of mass flow rate of solids on net energy consumptions was
studied. For these simulations, mass flow rate of fed air was changed proportional to the
solids mass flow rate. The mass flow rate of coal was then adjusted to get the same product
composition of solids (within ± 1%) at the kiln outlet. The predicted results in the form of net
energy consumption and percentage fill for different solids flow rates are shown in Figure
3.8a. As expected, as solids flow rate increases, the percent fill at kiln entrance increases. It
can be seen that the net energy consumption per unit weight of product decreases as solids
flow rate increases. This is because the net energy loss from the kiln does not increase
proportional to the solids mass flow rate. Thus it is beneficial from the point of view of
energy consumption to operate kiln with higher solids flow rate. Other operational concerns
like increase in dusting and mixing however need to be considered while identifying
maximum solids flow rate for any specific rotary cement kiln. The studies of flow, mixing
and heat transfer in transverse plane (for example, Khakhar et al., 1997 and references cited
therein) are useful to understand issues related to such operational concerns.
70
600 0.25
560 0.2
Energy Consumption,
Fill fraction at solid inlet

kJ/ kg clinker
520 0.15
480 0.1
440 0.05
400 0
20 30 40 50 60 70
Solid flow rate, kg/s
Figure 3.8a: Effect of solid feed rate on net energy consumption in rotary kilns
Influence of kiln rotational speed and extent of excess oxygen over stoichiometric
requirement on net energy consumption were also studied. The effect of kiln rotational speed
on the performance of kiln is shown in Figure 3.8b. It can be seen that as kiln RPM
decreases, the net energy consumption decreases. As the kiln RPM decreases the residence
time of solids in the kiln increases. For a fixed outlet composition the energy required to raise
temperature of solids to reaction temperature and energy required for clinker reactions is
more or less fixed. The heat losses also do not change significantly as RPM is changed.
However, since solid spends more time in the kiln resolidification of melt is more in kilns
rotated at lower RPM’s. Hence overall energy required for solid melting is less in kilns
operated at lower RPM. Thus, as long as adequate mixing of solids is taking place, it is
beneficial to operate the kiln at lower RPM. The simulations of excess oxygen indicated that
(not shown here for the sake of brevity), it is beneficial to operate the kiln with a slight excess
of oxygen (~1-4 %). Use of excess oxygen beyond 4 % leads to slight increase in the net
energy consumption per kg of clinker. Thus the model can be used for steering the design and
operation towards minimizing energy consumption per kg of clinker.
71
510
500
Energy consumption,
490
kJ/kg clinker
480
470
460
450
2 3 4 5 6 7 8 9
Kiln RPM
Figure 3.8b: Effect of RPM on net energy consumption in rotary kilns
The computational model developed in this work was also used to explore some
unconventional means to reduce energy consumption in kilns. For these simulations
freeboard model equations were not solved. Instead a temperature profile was assumed for
the gas phase and model equations for bed were solved (See section 3.1). The motivation of
this study was to explore for kiln energetics by possibly manipulating temperature profiles in
the freeboard region. The shape of the gas profile was assumed to be same as shown in Figure
3.9. This shape seems to be consistent with industrial observations and published results. The
position of maximum temperature was varied from 0.2 times kiln length to 0.8 times kiln
length. The value of maximum temperature at these positions was adjusted in such a way that
solids composition at the kiln outlet is same for all the cases. While carrying out these
simulations, the gas inlet temperature was coupled with the solids outlet temperature by
assuming a clinker cooling efficiency, η, as:
mG C pG (TG ,in − 300 ) = η m B C p TB ,out (27)
where m is mass flow rate, Cp is heat capacity, η is the cooler efficiency and T is temperature.
Subscripts G, B, in and out denotes gas, solids, inlet and outlet respectively. The temperature
72
2400
Area-weighted gas temperature (K)
2200
2000
Temperature, K
1800
1600
1400
1200
0 5 10 15 20 25 30 35 40 45 50
Kiln Length, m
Figure 3.9: Gas temperature profiles in freeboard region
of the gases leaving the kiln from the solid entrance was calculated by solving overall energy
balance on the kiln. Following procedure was use to simulate performance of a kiln.
The coal flow rate to the kiln (energy supplied to kiln in the form of coal, Ecoal) and the
maximum flame position were set to specific values. For these simulations the value of Ecoal
was fixed. Assuming the entrance and exit gas temperatures, the magnitude of the maximum
flame temperature was adjusted to get same (within ± 1 %) solids composition at the kiln
outlet. The inlet gas temperature was recalculated based on overall energy balance and
simulations were carried out with this new value of gas inlet temperature. This procedure was
continued till the inlet gas temperature did not change by ± 1 % of the previous value. Once
inlet gas temperature converged, outlet gas temperature was calculated based on overall
energy balance on the kiln. The value of gas outlet temperature was compared with the
initially assumed value and the whole procedure was repeated till the exit gas temperature
also did not change by ± 1 % of the previous value. Simulations to study effect of changing
position of maximum flame temperature were carried out for different values of Ecoal. The
simulated results are discussed in the following.
73
Typical predicted results for the net energy consumption in the kiln for varying position of
maximum flame temperature are shown in Figure 3.10. Corresponding numbers for net
energy consumption in kilns and position of coating formation with respect to maximum
flame position are tabulated in Table 3.9. As the maximum flame position moves towards the
solid inlet, coating formation takes place earlier in the kiln (See Table 3.9). Since coating is
formed early in the kiln, heat losses from the kiln are reduced due to additional resistance of
coating for heat transfer. The energy required by solids for clinker reactions and melt
formation remain more or less same. Therefore, as the maximum flame position moves
towards the solid inlet the net energy consumption in the kiln decreases. The computational
model and the results presented here will be useful to guide further work on modeling of
burner/ free board region and to re-design temperature profiles within the kiln to minimize
energy consumption. However, other concerns including possible operational difficulties
need to be investigated further in order to identify optimum yet practicable model of
operation for rotary cement kiln.
560
Ecoal = 924 kJ/kg clinkerl

Overall energy consumption kJ/kg clinker
540
Ecoal = 1050 kJ/kg clinker
Ecoal = 1176 kJ/kg clinker
520
500
480
460
440
0 0.2 0.4 0.6 0.8 1
Maximum flame position
Figure 3.10: Energy consumption in industrial kiln 1 for varying flame position
74
Table 3.9: Total energy consumption for industrial kiln 1 (kJ/kg clinker)
Maximum Flame Occurrence of Ecoal = 924 Ecoal = 1050 Ecoal = 1176
Temperature melt in kiln kJ/ kJ/ kJ/
Position (Normalised Kg clinker kg clinker kg clinker
from solid end)
0.2 0.52 473.59 473.39 472.15
0.4 0.58 486.52 484.92 483.37
0.6 0.66 505.09 502.19 501.08
0.8 0.77 540.56 540.74 538.94
3.3.5 Use of a secondary shell

Energy savings per kg of clinker by manipulating key parameters as discussed above are
possible to a rather limited extent (savings of about ~ 40 kJ/kg of clinker). There is however a
significant potential to reduce net energy consumption by using a secondary shell having low
surface emissivity and thermal conductivity around the kiln as proposed by Engin and Ari
(2005). The secondary shell could be covered with some insulating material to further reduce
heat losses. It should however be noted that reduction in losses from the kiln shell would
significantly increase temperature of kiln walls. Engin and Ari (2005) have not discussed this
issue in their paper. It is therefore essential to devise some way to ensure that the kiln shell
temperature does not rise beyond the acceptable limit. In this work, we propose to achieve
this by feeding air to the annular space between the kiln shell and the secondary shell. The
computational model was used to simulate performance of rotary kilns with secondary shell.
Simulations were carried out for the two values of gap between kiln shell and the secondary
shell: 0.1 m and 0.2 m. It was observed that the kiln shell temperatures are mainly controlled
by the mass flow rate of air. The simulated results of the gap of 0.1 m are shown in Figure
3.11. For these simulations the thickness of secondary shell used was 0.025 m (same as kiln
shell), thickness of insulation and thermal conductivity of insulation were set to 0.1 m and
0.05 W/m K respectively (Engin and Ari, 2005)
In absence of a secondary shell, the net energy loss to the surrounding was ~ 150 kJ/kg
clinker. With secondary shell and no gaseous stream in annular gap, this loss reduced to ~ 10
kJ/kg clinker. However, at such conditions, the temperature of kiln shell increases beyond
1000 K (see Figure 3.11). Operating cement kilns at such high shell temperatures is not
possible. It is therefore essential to reduce the kiln shell temperature by feeding gaseous
75
stream. As the mass flow rate of air increases, the temperature of kiln shell decreases. It can
be seen that the air mass flow rate of about 30 kg/s or more was required to bring the
maximum temperature below 650 K. As the mass flow rate of air increases, the extent of
energy recovery increases. It is however important to note that though the extent of energy
recovered increases with the mass flow rate of air through the gap, the quality (outlet
temperature) of the recovered energy decreases (see Figure 3.12). Appropriate balance of
extent of recovery and the outlet temperature of hot air stream from the gap needs to strike to
enhance the overall performance.
1600
Mair = 0 kg/s
Mair = 10 kg/s
1400
Mair = 20 kg/s
Mair = 30 kg/s
1200 Mair = 40 kg/s
Without secondary shell
Shell temperature, K
1000
800
600
400
200
0
0 0.2 0.4 0.6 0.8 1
Figure 3.11: Predicted profiles of temperature of kiln shell with secondary shell
If the recovered energy (at temperatures of about 2000 C) can be utilized within the cement
plant (refrigeration, drying of fly ash and so on), the use of secondary shell appears to be
promising for reducing net energy consumption.
The models and results presented here can be used to understand the kiln performance and to
possibly design appropriate secondary shell to reduce energy consumption. The
76
computational models can also be used to develop a rotary cement kiln simulator for training
and for optimization.
150 650
140 600
130 550
Exit air temperature, K

Energy recovered,
kJ/ kg clinker
120 500
110 450
100 400
90 350
80 300
0 5 10 15 20 25 30 35 40 45
Mass flow rate air, kg/s
Figure 3.12: Magnitude and quality of recover energy with the secondary shell
3.4 Conclusions
A comprehensive framework based on one-dimensional models was developed to simulate
complex processes occurring in rotary cement kilns. Models for simulating reactions and heat
transfer in the bed region and coal combustion and heat transfer in the freeboard region were
developed. Melting of solids and coating formation within the kiln was accounted. Coal
combustion in freeboard region was modeled by considering devolatilisation, gas phase
combustion and char reaction.
The model was used to simulate performance of three industrial kilns. The model predictions
agreed reasonably with the available plant data. The model was also able to capture influence
of key design and operating parameters on kiln performance. The model was used to evaluate
possible ways of enhancing performance of rotary cement kilns. The simulations indicated
77
that the operation of kiln with higher solid loading and lower RPM is favorable from the
point of view of energy consumption per kg of clinker. The use of a secondary shell with air
stream passing through the gap between the kiln shell and the secondary shell appears to be a
promising method to reduce energy consumption in rotary cement kilns. It is possible to
recover ~ 80 % of the energy lost to the surroundings (typically 150 kJ/kg clinker) by
judicious choice of secondary shell and air flow rate without degrading the quality of
recovered energy significantly.
78
4. Applications of Reaction Engineering based Models to clinker Manufacture
Abstract
This chapter presents an integrated reaction engineering based mathematical model for
clinker formation in cement industry. Separate models for pre-heater, calciner, and cooler
were initially developed. The model for pre-heater cyclones was formulated by assuming
solids and gas to be completely back mixed. The heat transfer coefficient between particle
laden gas and walls cyclones was calculated using empirical correlation. The model for
calciner was formulated based on Eulerian-Lagrangian approach. The coal particles and
raw meal particles were considered as discrete phases and gas phase was assumed to be
completely back mixed. Clinker cooler was modeled by assuming solids and gas in plug flow.
The heat transfer between solids and air in clinker cooler was predicted via empirical
correlations. The individual models were validated wherever possible. The models for pre-
heater, calciner, kiln (developed in previous chapter) and cooler were coupled with each
other via mass and energy communication through common boundaries to develop a coupled
simulator which was solved iteratively. The iterations were continued till the temperature of
solids and gases at exit of individual modules converged within error of one percent. The
model predictions agree well with the observations and experience from cement industry.
Several ways for reducing energy consumption were computationally investigated which
provide insights about influence of operating conditions on energetics of cement plant. An
easy to use, graphical user interface (GUI) based software called RoCKS (Rotary Cement
Kiln Simulator) was developed based on the integrated modules of pre-heater, calciner, kiln
and cooler. The integrated model, the developed software RoCKS (for Rotary Cement Kiln
Simulator) and results presented here will be useful for enhancing our understanding and for
enhancing the performance of clinker manufacturing
Publication based on this work

• Kaustubh S. Mujumdar, Ganesh K.V., Sarita B. Kulkarni and Vivek V. Ranade, “Rotary
Cement Kiln Simulator (RoCKS): Integrated Modeling of Pre-heater, Calciner, Kiln and
Clinker Cooler”, Chemical Engineering Science, 62, 2007, 2590-2607.
79
4.1 Introduction
In the previous chapters of this thesis, reaction engineering based computational model for
cement kilns was developed. It was shown that such models not only helped in developing
better understanding for cement kilns but also provided some useful clues for reducing
energy consumption. However, as described in chapter 1, in a clinker manufacturing process,
the pre-heater, calciner, kiln and cooler are strongly coupled with each other. The input to
rotary kilns in terms of mass fraction and composition can vary if operating conditions in
calciner or cooler are varied. Therefore, the utilization of the model developed for the kiln
would be rather restricted if it is not coupled with computational models for pre-heater,
calciner, and cooler. It is therefore essential to develop an integrated model for pre-heater,
calciner, kiln and cooler in order to capture key characteristics of clinker manufacturing and
to enable the model to be used as simulation or optimization tool. Such an attempt is made in
this chapter. The motivation of the present work was to develop a reliable frame work of
reaction engineering based computational model for cement industry and use this model
subsequently for possible exploration of performance enhancement. Based on this integrated
model a simple graphical user interface based software called RoCKS (Rotary Cement Kiln
Simulator) was also developed. The overall organization of this chapter is briefed below.
The key issues that needs to be considered while developing individual models are discussed
initially in section 4.2. The computational model and the modeling strategy are thereafter
presented in section 4.3. Section 4.4 reports the results of computational simulations of
integrated model with respect to key operating parameters. Thereafter the computational
model was utilized to gain insights about cement plant energetics via numerical experiments.
Key findings of the study are summarized at the end.
4.2 Key Issues and Modeling Approach
Various processes occurring in pre-heater, calciner, kiln and cooler need to be adequately
considered while developing computational model for cement plant. Key issues governing the
performance of individual units are schematically shown in Figure 4.1. Initially the key issues
of individual equipments and review the previous work related to it are discussed to provide
background for the model developed in this work.
80
Pre-heater Calciner
Complex hydrodynamics, Complex hydrodynamics,
gas-solids heat transfer (Gas, coal combustion, turbulent gas
solids, walls), particle laden phase combustion, calcination,
turbulent flow, losses heat transfer (Gas, solids,
walls), Losses
Strongly
Coupled
Rotary Kiln
Coal/gas phase
combustion, bed height Clinker Cooler
variation , clinkerization reactions,
Gas solids heat transfer,
melting/solidification, coating
temperature gradients in bed
formation, radiation,
re-solidification of melt
bed-freeboard exchange
losses
losses
Figure 4.1: Key issues in modeling cement plant
4.2.1 Cyclone preheaters

The calcineous raw meal is passed through a set of pre-heater cyclones (depending on pre-
heater assembly) before it is sent to the calciner. In the pre-heater section, the raw meal is
pre-heated to around calcination temperature by hot gases coming from the calciner. The
operation of cement pre-heater is similar to that of a conventional cyclone. Solids are fed to
the cyclone along with the gas coming from the previous pre-heater. Both solid and gas
spiral down towards the cyclone bottom where the gas reverses its direction and leaves
through the exit duct while the solids leave from the bottom to the next unit. The flow inside
the cyclone is characterized by high swirl and turbulent motion. This provides excellent heat
transfer between gas and solids. We used computational fluid dynamics (CFD) based model
for studying mixing and heat transfer of gas solid flows in cyclones. Our CFD simulations
(carried out for typical values of operating conditions as gas flow rate = 60.83 kg/s; solid
flow rate = 50 kg/s; inlet gas temperature = 740 K; inlet solid temperature = 500 K; cyclone
diameter = 6 m; height of cyclone = 11 m) indicated that the solid and gas temperatures
inside the cyclone were quite uniform (in range of 637 K and 644 K) and close to exit
temperatures (average temperature of 640 K). Therefore, in this work, each pre-heater
cyclone was treated as completely mixed cell for individual phases and was therefore
represented by a pair of temperature (one for gas phase and the other for solid phase). The
81
heat losses from cyclone are controlled by the heat flux across the cyclone walls. Therefore it
is essential to predict the heat transfer between particle laden gas and cyclone walls in the
pre-heater cyclone. The empirical relation proposed by Gupta and Nag (2000) was used to
determine these losses.
4.2.2 Calciner
The operation of calciner is also like a conventional cyclone. The last cyclone in the preheater
assembly generally acts as a calciner. Therefore some of the key issues that mathematical
model needs to capture in calciners (for example, hydrodynamics and heat transfer between
particle laden gas and walls) are essentially same as that in the preheaters. Therefore we
assume gas phase to be completely back mixed and empirical correlation to model heat
transfer from particle laden gas to cyclone walls. The main difficulty in modeling cyclone
calciner is to estimate the residence time of raw meal in the cyclone. Few empirical
correlations have been proposed to predict the average residence time of particles in cyclones
(Kang et al., 1989; Lede et al., 1987). However all the attempts for prediction of residence
time in cyclones were for lab scale cyclones and non of the studies were reported for
industrial scale cyclones. Moreover, the proposed empirical correlation works only for their
own experimental data and do not work if applied to other system. Due to unavailability of
any information for particle residence time in the cyclones as a first approximation this
parameter was treated as adjustable parameter in the model. The raw meal residence time in
the calciner was adjusted to get the desired degree of calcination the calciner.
Other than this, it is also essential to obtain relevant kinetics for calcination reaction in
calciner. Thermal decomposition of limestone calcination is a complex process. Predicting
kinetics of calcination reaction has been subject of an extensive research over the years.
However, still there is no reliable consensus on the calcination kinetics. It is well established
that the overall reaction rate is function of the following three rate processes, 1) heat transfer
(to the surface and then to the reaction zone) through the particle 2) mass transfer (both
internal and external to the surface) and 3) chemical reaction at the interface of
undecomposed particle and product. Relative rates of these three transport processes depend
on particle size, particle morphology and operating conditions and varies with varying
operating conditions. Hence, a wide discrepancy is observed in the proposed rates for
calcination reaction. It was also reported that the proposed models fit adequately their own
experimental data. However, when they are applied to other results most of them fail (Khinast
82
et al., 1996). As shown in Chapter 3 in this thesis, we have compared models proposed by
about 18 investigators on available experimental data. The data reported in experiments of
Watkinson and Brimacombe (1982) was chosen for this study. Watkinson and Brimacombe
(1982) have reported experimental data on calcination of limestone in experimental kiln. The
experimental conditions of their experiments were close to industrial operations (bed
temperature ~ 1000 to 1300 K) and therefore were used to find calcinations kinetics in this
work.
4.2.3 Rotary kiln

The partially calcined raw meal is passed slowly to rotary kiln where the clinkerization
reactions occur. In the initial part of the kiln remaining calcination occurs. Other solid-solid
and solid –liquid clinkerization reactions take place as solid bed moves towards the burner.
Part of the solids melts in the kiln. The melt formation causes an internal coating on kiln
refractories. Counter current flow of gas entrains solid particles in the free board region. Such
entrainment enhances rates of radiative heat transfer by increasing effective emissivity and
conductivity. In a nutshell, the main key issues for modeling rotary cement kilns are
estimating residence time of solids in the kiln, cinkerization reaction in bed region, coal
combustion in freeboard region, heat transfer between bed freeboard and walls,
melting/coating formation around the kiln walls. These issues were discussed in detail in
Chapter 3 and therefore are not repeated here.
4.2.4 Clinker cooler

The hot solids from the kiln are discharged on the grate of clinker cooler. As the grate moves
with uniform speed along the cooler length, solids lose their heat to cross flow air. A part of
air is generally sent to the kiln as secondary air, a part to calciner as tertiary air and part is
lost to the surroundings (vent air). The most important key issue in modeling grate coolers is
predicting the heat transfer coefficient between hot solids and cross flow air. Unfortunately
there is no information on modeling of heat transfer in such cases. In absence of any relevant
information we have used heat transfer correlation in packed bed reactors to estimate the heat
transfer. Nsofor and Adebiyi (2001) have carried experimental measurements and presented
correlation for forced convection gas particle heat transfer coefficient at high temperatures
(200 C to 1000 C). Since the temperatures in clinker cooler are in the same range this
correlation was used to model heat transfer coefficient between solids and gas. The heat
83
transfer coefficient was coupled with an efficiency factor which was adjusted to match the
temperatures of secondary and tertiary and coming out of cooler.
The key issues for preheater, calciner, kiln and cooler presented in the above section suggests
that in order to develop a reliable computational models for cement industry it is required to
incorporate all the above-mentioned processes occurring in preheater, calciner, kiln and
cooler. Moreover, appropriate algorithms methodologies need to be developed for coupling
of individual models. Such an attempt is made here. The motivation of the present work was
to predict various processes occurring in preheaters, calciner, kiln and clinker cooler
simultaneously in a single framework. The various physical processes considered during
development of the model are discussed in the following section.
4.3 Computational Models and Solution Methodology
Individual sub models for pre-heater, calciner, and cooler are discussed in the following
section.
4.3.1 Pre-heater model
A schematic of pre-heater unit considered for developing computational model is shown in
Figure 4.2. The present framework of computational models was developed for dry process
of clinker formation since this process is widely used in cement industry.
Mg,Mse,i, Ti
Ms,i-1, Ti-1
Tiw, i Tow, i
i 12
Radiation and convection
Mg, Mse, i-1, Ti-1 Losses
1. Refractory
2. Shell
M s, i, T i
Figure 4.2: Schematic of Pre-heater
84
For the dry processes, the moisture content is generally present in very small amount
[typically ~ 0.5%, see for example Engin and Ari (2005), Peray (1986)]. The energy
requirements for removing the moisture from the feed being small (less than 0.5% of the total
energy consumption), the feed was considered to be free of moisture in this work. However,
the developed framework is quite general and including evaporation of moisture from the
feed is straightforward.
In Figure 4.2, Ms, is the mass of solids entering the cyclone. Mg is the mass of the air entering
the cyclone. Mse is the mass of solids entrained from a cyclone. Tiw, i and Tow, i represent the
inner and outer wall temperature for the ith cyclone. Each cyclone was assumed to be lined
with refractory of thickness tr
Thus, for any ith cyclone in pre heater assembly following are the inlet streams
1. Solids from the (i-1)th cyclone (Ms, i-1 at temperature T i-1).

2. Solids that are entrained by gas from (i+1)th cyclone (Mse, i+1 at temperature Ti+1).
3. Air from (i+1)th cyclone (Mg at temperature Ti+1).
The outlet streams for this cyclone are:
1. Solids going out of cyclone (Ms, i at temperature Ti).

2. Solids that are entrained by gas (Mse,i at temperature Ti).
3. Air going out (Mg at temperature T i+1).
The steady state material balance equation for ith cyclone was written as
Ms,i-1 + Mse,i+1 = Ms,i + Mse,i (4.1)

Mse,i = (1-ηm, p) × Ms,i (4.2)
In the above equations ηm, p represents the mass efficiency of the ith cyclone. M represents the
mass of the solids (in kg/s) and subscripts s and se represent solids and entrained solids
respectively as explained earlier.
The steady state energy balance for the ith cyclone was written as
M s ,i-1 × C p, s × Ts, i-1 + M se, i +1 × C p,s × Tc, i +1 + M g × C pg × Tc, i +1 = M s, i × C p,s × Tc , i + M se , i × C p,s × Tc , i

[
+ M g × C p,g × Tc, i + h cyc × Acyi × Tc , i − Tiw, i ] (4.3)
In the above equations Cp,s and Cp,g represents the specific heat of solids and air respectively.
Subscript g represents the air and Tc represents the temperature of solids and air in the ith
cyclone. hcyc represents the heat transfer coefficient for energy exchange between particle
85
laden gas and cyclone inner walls. hcyc was evaluated from empirical correlation given by
Gupta and Nag., (2000) for heat transfer in cyclones given by
hcyc d c ∆P ∆P
= 702 .818 + 9 .0287 × 10 −14 u o Re + 11 .1385 ( 2
) + 4.50398 × 10 −5 ( ) Re + R c
kg u o uo
⎛ T 4 iw − T 4 g ⎞ dc
Rc = σF p − w ⎜ ⎟
Where ⎜ T −T ⎟k (4.4)
⎝ iw g ⎠ g
The term η h, p is an adjustable parameter (1 in the present simulations) that is provided in the
model which can be used to fine tune the model to fit it with industrial data. Acyi (m2) is the
internal surface area of the cyclone. The LHS of Equation (4.3) thus represents the total heat
entering the cyclone and RHS represents the heat leaving out of the cyclone. At steady state
the energy given to cyclone walls must be same as heat conduction in through refractory and
cyclone walls, which is equal to loss from shell walls due to convection and radiation. The
energy balance for heat transfer in cyclone cross section was written as:
[
h cyc ⋅ Acyi ⋅ Tc , i − Tiw, i ] =
[
2 ⋅ π ⋅ L ⋅ k r ⋅ Tiw, i − Tr , i ] (4.5)
⎛r ⎞
ln⎜⎜ r ⎟⎟
⎝ ri ⎠
[
2 ⋅ π ⋅ L ⋅ k r ⋅ Tiw, i − Tr , i ]= [
2 ⋅ π ⋅ L ⋅ k s h ⋅ Tr , i − Tow, i ] (4.6)
⎛r ⎞ ⎛r ⎞
ln⎜⎜ r ⎟⎟ ln⎜⎜ 0 ⎟⎟
⎝ ri ⎠ ⎝ rr ⎠
[
2 ⋅ π ⋅ L ⋅ k s h ⋅ Tr , i − Tow, i ]= h [
⋅ Acyo ⋅ Tow, i − T0 ] [
+ σ ⋅ ε cy ⋅ Acyo ⋅ Tow, i − T0
4 4
]
conv
⎛r ⎞
ln⎜⎜ 0 ⎟⎟
⎝ rr ⎠
(4.7)
In the above equations, Tiw, i represents the internal wall temperature of the ith cyclone, Tr,i
represents the temperature of interface of refractory and shell, Tow, i represents the
temperature of external wall of the ith cyclone and T0 represents the ambient temperature. L
represents the total height the cyclone, kr represents the thermal conductivity of the refractory
and ks represents the thermal conductivity of cyclone walls. r0 represents the external
diameter, rr represents the internal diameter of the shell and ri represents the cyclone internal
86
diameter. Acyo represents the external surface area of the cyclone. hconv represents the
convective heat transfer coefficient of external wall (taken as 30 W/m2K as presented in
Chapter 3), εc is the emmissivity of the cyclone outer wall and σ is the Stefan-Boltzman
constant for radiative heat transfer.
The above material and energy balances were written for n cyclones. For each cyclone there
are 6 equations and 6 unknown variables. Therefore for n cyclones there are 6 × n variables
number of variables and 6 × n equations. These set of non linear algebraic equations were
solved using Newton-Rapson method to get mass and temperature of solids, gas and walls
walls for preheater cyclones
It was essential to test the framework of models developed for pre-heaters at least
qualitatively before implementing it in the integrated model. To do this the developed
computational model for pre-heater cyclones was used to predict the temperatures and energy
losses for two different scenarios. The first test case considered 3 cyclones in pre-heater
assembly, while, in the second test case 4 cyclones were considered in the pre-heater
assembly. All the cyclones considered in the present study were of identical dimensions
(Internal diameter 3 m; External diameter 3.05 m; height 5 m). Calcineous raw meal flow rate
to first cyclone was set to 50 kg/s at a temperature of 300 K and air flow rate from the
calciner was set to 50 kg/s at temperature of 1100 K. for both the test cases. The flow rates
and temperatures of air and raw meal fed to cyclones was obtained from industrial data. The
efficiency of each cyclone (i.e. mass collected from cyclone bottom to mass fed) was
assumed to be 99.99 % from experimental measurements of Gupta and Nag (2000). Gupta
and Nag (2000) have also reported empirical correlation heat transfer coefficient for particle-
laden gas in cyclones. This correlation was used to calculate heat transfer coefficient between
particle laden gas and cyclone walls. Two pre-heater units comprising four and three cyclones
were simulated. The predicted results are shown in Figure 4.3a and 4.3b and are summarized
in Table 4.1.
87
1300 1300
Solid Temperature Solid Temperat ure
Gas Temperature Gas Temperat ure
1100 1100
Decreasing gas t emperat ure

900
Temperature, K
900 Decreasing gas temperature
Temperature, k
700 700
500 500
Increasing solid t emperat ure

300 Increasing solid temperature 300
Solid Inlet
Solid Inlet
100 100
0 1 2 3 4 0 1 2 3
Cyclone number Cyclone number
(a) 4 cyclone case (b) 3 cyclone case
Figure 4.3: Predicted results for the pre-heater units
Table 4.1: Predicted cyclone temperatures and loss
Four Cyclone Case Three Cyclone Case

Temperature, K Temperature, K
Cyclone (Solid and gas at exit) Loss (kW) (Solid and gas at exit) Loss (kW)
1 434 283.81 479 383
2 574 606.50 666 833
3 726 992.95 869 1389
4 898 1475.55 - -
It can be seen from Figure 4.3 and Table 4.1 that with an additional cyclone in the assembly,
there is more heat exchange between solids and gas i.e. outlet temperature of solid from the
last cyclone increases while outlet gas temperature decreases. Thus, the temperature of solids
coming out of pre-heater assembly considering 4 cyclones is more than one considering 3
cyclones, which seems to be reasonable. However, a 3 cyclone assembly has about 25%
lower losses due to lower surface area. The model can thus be used to evaluate alternative
pre-heater configurations for different pre-heater assemblies.
88
4.3.2 Calciner model
The mathematical model of calciner was based on a schematic shown in Figure 4.4. The
following assumptions were made:
1. Gas phase is completely back mixed.
2. The raw meal and coal particles were treated as discrete phases having uniform
particle size. All the particles were assumed to have the same residence time in
calciner.
The mass and energy balance equations are presented below.
4.3.2.1 Discrete phase

Coal particles
Mass balance for coal particles in calciner was written as:
− E1
dm p ,c
= − Ao ⋅ f v ,o ⋅ m pc ,o − k s1 ⋅ e RTc c
A p ⋅ pO 2 . (4.8)
dθ
Gas out
Coal in
Air in
Loses
Partially calcined raw meal
Figure 4.4: Schematic of cyclone calciner
89
In the above equation, A0 is the devolatilization constant, fv,0 is the initial mass fraction of
volatiles in coal particle and the mpc,0 is the initial mass of coal particle ,ks1 is the rate
constant of char (1 kg/m2-s-kPa, Hamor et al., 1973) , cAp is the available surface area of
particle, E1 is the energy of activation (6.81E4, Hamor et al., 1973) and po2 is the partial
pressure of O2 . θ is time spent by a coal particle in calciner. The first term in the RHS of
above equation represents the loss in discrete phase mass due to devolatilization of coal
particles and the second term in the RHS of above equation represents the loss in discrete
phase due to char reaction
The individual species balance for rate of mass change due to devolatilization is given by,
d (m p ,v )
= − Ao ⋅ f v , o ⋅ m pc , o (4.9)
dθ
Where, mp,v is the mass of volatiles in coal particle.
The individual species balance for rate of mass change due to char combustion is given by
− E1
d ( m p , char )
= − k s1 ⋅ e RTc ⋅c Ap ⋅ pO 2
dθ (4.10)
where mp,char is the mass of char.
The energy balance for a coal particle is given by

dm p , c C p , cTc c
= hconv ⋅c A p ⋅ (Tg − Tc ) + σ ⋅ ε c ⋅c A p ⋅ (Tg − Tc ) + f c ⋅ ∆ H comb ⋅ rcomb
4 4
dθ
dm p , c
+ C p , c .Tc .
dθ
(4.11)
In the above equation Cp, c is the specific heat of coal and TC is the coal temperature. Coal
was assumed to enter the kiln at 300 K. Tgout is the gas temperature. ∆Hcomb is the heat
released due to coal combustion and fh is the fraction of energy released due to coal
combustion, which is absorbed by the particle. This factor was set to 0.3 following the
recommendation of Boyd and Kent (1986). The first and second terms in RHS of above
equations are the energy absorbed by the coal particle due to convection and radiation from
the gas phase during the heat up. The heat transfer coefficient, hc, is evaluated using the
correlation of Ranz and Marshall (1952) as:
90
hc d p
= 2 + 0.6(Re) 0.5 (Pr) 0.33 (Re up to 104; Pr>=0.7) (4.12)
kg
The final term in Equation (4.11) represents the loss of energy from coal particle due to loss
of mass.
Raw meal particle

The calcination reaction is given as
CaCO 3 → CaO + CO 2 (4.13)
The over all mass balance for the raw meal particle is given by,
d Lm p Mwco 2
= − rcalc .
dτ Mwcaco 3
(4.14)
where
− E2
rcalc = k s ⋅ e
r RTp
⋅ 4 ⋅ π ⋅ r p ⋅ Mw caco3
2
(4.15)
where, the Lmp , Mwco2, Mwcaco3 are mass of raw meal particle, molecular weight’s of carbon
dioxide calcium carbonate respectively is the rate of calcination of calcium carbonate. rcalc is
the rate of calcination of calcium carbonate, rp radius of shrinking raw meal particle
respectively. The term rks is the rate constant of calcination of calcium carbonate of
calcination. It has been well established that presence of CO2 in gas phase inhibits calcination
rate (Stanmore and Gilot, 2005). However, the nature of function in which the rate is affected
is again a subject of disagreement. In this work we use a simple linear form of relationship to
model effect of presence of CO2 on rate of calcination reaction as
− E2
rcalc = k s′ ⋅ e RTp ⋅ 4 ⋅ π ⋅ r p ⋅ Mw caco3

2
r
(4.16)
where
r
k s′ = r k s PCO2 < 10 −2 Peq
⎛ PCO 2 − Peq ⎞
r
k s′ = r k s ⎜ ⎟ 10 − 2 Peq < PCO2 < Peq (4.17)
⎜ Peq ⎟
⎝ ⎠
−19680
( )
and Peq = 1.826 × 10 7 e T
(4.18)
The individual species balances for limestone and calcium oxide are given as,
91
d LmCaCO 3
= − rcalc
dτ (4.19)
d LmCaO rcalc .Mwcao

= (4.20)
dτ MwCaCO 3
The energy balance for the raw meal particle is given as,
d ( L m p C p , sT p ) d Lm p
= hconv ⋅ A p ⋅ (Tg − T p ) + σ ⋅ ε L ⋅ A p ⋅ (Tg − T p ) + C p , sT p
L L L 4 4
dτ dτ
(4.21)
where, Tp is the temperature of raw meal, εL is the emissivity of solid particle, LAp is the area
of the raw meal particle and τ is residence time of raw meal particle in the calciner. Lhconv was
estimated as per Equation (4.12)
4.3.2.2 Continuous phase
The over all gas mass balance is given as,
dm g
= m gin − m gout +[ c m pin − c m pout ]⋅c N p +[ L m pin − L m pout ]⋅ L N p
dt
(4.22)
where, mg is the mass of the air, cNp is the number of particles of coal coming in per unit time
and LNp is the number of particles of raw meal coming in per unit time. The individual
species mass balance for rate of change of mass of oxygen, carbon dioxide, volatile matters
and water can be written as:
dy o2 1 [ m p ,c ⋅ y c ,cin − m p ,c ⋅ y c ,cout ]⋅ c N p ⋅ Mw O 2
= ( m g ,in ⋅ y o2 in − m g ,out ⋅ y o2 out −
dt mg Mw char (4.23)
rcombg dm g
−( ) ⋅ V react ⋅ Mw o 2 ⋅ Z O 2 − y o2 ⋅ )
Mw vol dt
dy CO2 1 [ m p ,c ⋅ y c ,cin − m p ,c ⋅ y c ,cout ]⋅ c N p ⋅ Mw CO 2

= ( m gin ⋅ y CO2in − m gout ⋅ y CO 2 out + +
dt mg Mw char
rcombg dm g
[ m p , Lin − m p , Lout ]⋅ L N p +( ) ⋅ V react ⋅ Mw CO 2 ⋅ Z CO2 − y CO 2 ⋅ )
Mw vol dt
(4.24)
92
dyv 1
= ( mg ,in ⋅ yv ,in − mg ,out ⋅ yv ,out + [ m p ,cin ⋅ yv − m p ,cout ⋅ yv ]⋅c N p
dt mg
rcombg dmg
−( ) ⋅Vreact ⋅ Mwvol ⋅ Z vol − yvol ⋅ )
Mwvol dt
(4.25)
dy w 1 rcombg dm g
= ( m gin ⋅ y win − m gout ⋅ y wout + ( ) ⋅ V react ⋅ Mw w ⋅ Z H 2O − y w ⋅ )
dt mg Mwvol dt (4.26)
where, yo2, yco2, yv, yw are the respective mass fractions of oxygen, carbon dioxide, volatile
matters and water. MwO2, MwCO2 , Mwvol and Mww are their respective molecular weights.
Vreact is the volume of reactor. Subscripts in and out represent the inlet and outlet conditions
and Z is the stoichiometric coefficient.
The energy balance equation for the gas phase is given as,
dm g C p , g T g
= m gin ⋅ C p , g ⋅ T g ,in − m gout ⋅ C p , g ⋅ T g + S gcomb + S ccomb + S calc
dt
− h cyc ⋅ Acyi ⋅ (T g − Tiw )
(4.27)
Where,
S gcomb = rcombg ⋅ H combg ⋅ V react (4.27a)
θc
d cm p
S ccomb = − ∫ ( hconv ⋅ A p 0 ⋅ ( T gout − T c ) + σ ⋅ ε c ⋅ A p 0 ⋅ ( T gout − Tc ) + C p .Tc .
c c c 4 4 c
0
dθ
− ( 1 − f c ) ⋅ H comb ⋅ rcomb ) ⋅ d θ (4.27b)
τr
S calc = − ∫ ( hconv ⋅ L A p 0 ⋅ ( T g out − T L ) + σ ⋅ ε L ⋅ L A p 0 ⋅ ( T g out − T L )
4 4
d Lmp (4.27.c)
+ C P TL
L
− rcalc ⋅ H calc ) ⋅ d τ
dτ
In the above equations, Sgcomb, Sccomb is the heat source term for gas-phase from volatile
combustion and char combustion respectively. Scalc is the heat sink term from calcination.
Hcombg, Hcalc are the enthalpies of volatile combustion and calcination. hcyc is the convective
93
heat transfer coefficient of the particle laden gas. The steady state equations across the
cyclone walls were written same as that of preheaters explained in the previous section to
obtain temperature of calciner internal walls, refractory and outer walls .
An iterative method was developed to solve the calciner model equations. The model
equations for gas phase were solved assuming steady state. For the first iteration, source
terms from discrete phase were assumed to be zero. The temperature and mass of species
obtained at calciner exit by solving continuous phase were used in discrete phase equations to
get the new source terms from the discrete phase. The sources from the discrete phases were
passed to continuous phase to get the new mass and temperature terms for discrete phase.
This procedure was continued till the subsequent changes in temperature of gas phase were
within ± 0.1 %. Suitable under-relaxation parameters were defined to accelerate convergence.
Typically about 20 iterations were required to achieve convergence. The differential
equations for discrete phase were solved by modified Gear’s method implemented in
ODEPACK (Hindmarsh, 1983). The algebraic equations for continuous phase were solved
using Newton-Raphson method.
Calculate gas inlet temperature
Solve Continuous Phase
Get gas temperature,

mass terms
Solve Discrete Phase
No Convergence Calculate Sources
Converged gas temperature
Figure 4.5: Solution Methodology
94
The computational model of the calciner was then further used to understand influence
particle diameters and coal flow rates on calciner performance. Some of these results are
discussed in the following. The mathematical model of calciner was evaluated by comparing
predicted results with data obtained from ACC. The schematic of the calciner is shown in
Figure 4.6 and details of input conditions provided are specified in Table 4.2. The data
obtained from ACC unfortunately did not specify the calorific value of coal and actual
amount of coal fed to the calciner. From the knowledge of total coal flow rate (which was
specified), the coal fed to the calciner was taken to be proportional to the extent of calcination
taking place in that calciner. Knowledge of residence time of solids in their calciner was also
not available precisely. Simulations were therefore carried out over a range of calorific values
of coal (4500 to 5200 kcal/kg coal) and of residence time of raw meal in the calciner (2-20 s).
The results obtained from these simulations are shown in Figure 4.7.
Mixture to Cyclone 5
for separation
3.91 m
6.00 m
Cyclone 4
material
Coal Burners
ary air duct Tertiary air duct
2.34 m
Kiln gases
Figure 4.6: Schematic of Kymore line-1 calciner (from Anathraman, 2006, private
communications)
95
Table 4.2: Data specified from Kymore line-1*
Sr. Variable Inlet Outlet
No.
1 Tertiary air mass flow rate, kg/s 39.33 -
2 Tertiary air temperature, K 1123 -
3 Kiln air mass flow rate, kg/s 27.33 -
4 Kiln air temperature, K 1293 -
5 Exit air mass flow rate, kg/s - 79
6 Exit air temperature, K 1168
7 Raw meal mass flow rate, kg/s 63.05 52.5
8 Raw meal temperature, K 1073 1168
9 Coal flow rate, kg/s 5
10 Mass fraction CaCO3 0.507 0.097
11 Mass fraction CaO 0.177 0.65
*
- (52% Calcination of raw meal at Calciner inlet; 92 % Calcination of raw meal at Calciner outlet)
100 1200
95
1180
Solid exit temperature, K
90
4600 kcal/kg coal
4700 kcal/kg coal
% Calcination
85
1160
4850 kcal/kg coal
80
4950 kcal/kg coal
1140
75 5150 kcal/kg coal
70
1120
65
60 1100
0 4 8 12 16 20 24 28
Particle residence time ,s
Figure 4.7: Simulation results for various coal calorific values and residence time of
solids
96
The simulation results indicate that for a coal with calorific value of ~ 4500 to 5200 kcal/kg
coal, to achieve 80-95 % conversion the raw meal spends about 4-12 sec in the calciner (see
window marked in Figure 4.7). This looks quite reasonable. The quantitative comparison of a
specific simulation result with industrial data is given in Table 4.3.
Table 4.3: Comparison of model prediction with industrial data at calciner exit
(Calorific value of coal = 4850 kcal/kg coal; Residence time of solids = 8 s)
Sr. Model Industrial
No Predictions data
1. Mass Fraction CaCO3 8.09 9.92
2. Mass Fraction CaO 65.47 63.86
3. Mass Fraction SiO2 18.62 18.46
4. Mass Fraction Al2O3 4.47 4.43
5. Mass Fraction Fe2O3 3.35 3.32
6. Temperature of solids, K 1160 1158
The model presented in the previous section was used to evaluate the performance of a
typical calciner. To test the framework of the developed model, the volume of calciner was
taken as 50 m3. The inlet coal flow rate to the calciner was set to 1 kg/s and raw meal inlet
flow rate was set to 35 kg/s. The inlet temperatures of coal and raw meal were taken as 350 K
and 1000 K respectively. The gas inlet temperature to the calciner (from kiln exhaust gas and
tertiary air from cooler) was set to 1194 K. The inlet conditions to the calciner were
reasonable when compared with available data. For the initial simulations we have assumed
the residence time of raw meal and coal particles inside the calciner (10 s). The framework is
general enough so that mathematical equation to evaluate residence time based on calciner
operating conditions can be incorporated later. The kinetic parameters for char combustion,
gas phase combustion and calcination reaction were same as used for kiln in chapter 2. With
97
90 1100
Raw meal Temperature ,K

1080
85
1060
% Calcination
80
1040
75
1020
70
1000
65
980
60 960
0 20 40 60 80 100
Raw Meal DP -microns
Figure 4.8a: Influence of particle size
80 1240
70
Raw meal temperature ,K

1200
60
1160
% Calcination
50
40 1120
30
1080
20
1040
10
0 1000
0 2 4 6 8 10 12 14
Residence Tim e (s)
Figure 4.8b: Influence of particle residence time
100 1150
90
Raw meal temperature ,K
80 1110
% Calcination
70
60 1070
50
40 1030
30
20 990
10
0 950
0 0.5 1 1.5 2
Coal Flow rate (Kg/s)
Figure 4.8c: Influence of coal flow rate

Figure 4.8: Predicted performance of a typical calciner
98
these input and kinetic parameters, the model predicted calcination of 74 %, exit gas
temperature of 1057 K, exit solid temperature of 1057 K and losses of about 102 kW (<1 %
of total energy input). The above values seem to be reasonable when compared qualitatively
with industrial observations.
The influence of variation in inlet raw meal particle diameter on percentage calcination and
temperature of raw meal at calciner exit is shown in Figure 4.8a. As expected, it can be seen
that with decrease in particle size the percentage of calcination increases. Figure 4.8b shows
the predicted influence of particle residence time on the performance of calciner. It can be
seen that the extent of calcination increases with increase in particle residence time. The
effect of change in coal flow rate on the calciner performance is shown in Figure 4.8c. Thus
the computational model developed in this work is useful to simulate behaviour of calciners
and to understand influence of key design and operating parameters on its performance. The
model needs to be validated by comparing predicted results with industrial data.
4.3.3 Kiln model
The computational model for rotary kiln was developed by coupling bed height model, bed
model and freeboard was explained in detail in chapter 3 and will not be discussed here.
4.3.4 Cooler model
The mathematical model of cooler was based on a schematic shown in Figure 4.9. Solids of
uniform particle size and constant porosity were assumed to move in a plug flow with
constant grate speed vg in x direction. Air was assumed to enter in a cross flow mode with
respect to solids in y direction. The cooler was divided into three sections as can be seen from
Figure 4.9. The amount of air fed in each section was distributed proportional to fraction of
length of each section. The air coming out of first section was assumed to be sent to the
secondary air. The air from the second section was assumed to go to the calciner as tertiary
air and air from third section was assumed to be lost to the surrounding as vent air. To get the
temperature profiles of solid bed and air, the clinker cooler was divided into n segments along
the length of the cooler (L) and m segments along the height of the cooler (H). Mass and
energy balances were solved for these segments. Conductive heat transfer was considered for
solids in both horizontal and vertical directions. Convective heat transfer coefficient between
99
Secondary air Tertiary air Vent air
Hot clinker, Ts, in
Cold clinker, Ts, out

Grate
Cooling Air, Ta
Figure 4.9 Schematic of grate cooler
Secondary air Tertiary air Vent air

⎛ ∂T ⎞
⎜⎜ ⎟⎟ = 0
⎝ ∂y ⎠
L
⎛ ∂T ⎞
⎜ ⎟=0
⎝ ∂x ⎠
Hot clinker, Ts, in Cold clinker, Ts, out
⎛ ∂T ⎞
⎜ ⎟=0
⎝ ∂x ⎠
⎛ ∂T ⎞
⎜⎜ ⎟⎟ = 0
⎝ ∂y ⎠
Cooling Air, Ta
Figure 4.10: Boundary conditions
air and solids was calculated from empirical correlation assuming solids as packed bed. The
boundary conditions used in the model are shown in Figure 4.10.
The model equations are presented in the following.

Mass balance for solids can be written as:
100
dm s ( i , j )
=0 (4.28)
dx
Assuming steady state operation, the energy balance equation can be written as
∂ ( ρ s (1 − ε ) u xs c sp T s ) ∂ ( ρ s (1 − ε ) u ys c sp T s )
+ =
∂x ∂y
⎧ ∂T s ⎫ ⎧ s ∂T
s
⎫ (4.29)
∂ ⎨ (1 − ε ) K s ⎬ ∂ ⎨ (1 − ε ) K ⎬
⎩ ∂x ⎭ ⎩ ∂y ⎭
+ − a h cv (T s − T a )
∂x ∂y
In this equation ρs is the cement clinker density, Cps is clinker heat capacity, uxs is grate
speed, and Ts is clinker temperature of solid at any point, Ks is clinker thermal conductivity,
a is the convection area factor between the clinker and air, hcv is convective heat transfer
coefficient between solid clinker and air, Ta is air temperature at any point in the cooler. In
Equation (4.29) the first and second terms of the right hand side represent the conduction.
The last term in right hand side represents convective heat transfer between the air and solids.
The mass balance for air can be written as:

dma ( i , j )
=0 (4.30)
dy
Energy balance for air can be written as:
∂ ( ρ a ε u xa c ap T a ) ∂ ( ρ a ε u ya c ap T a )
+ =
∂x ∂y
⎧ ∂T a ⎫ ⎧ ∂T a ⎫ (4.31)
∂⎨K aε ⎬ ∂ ⎨ K aε ⎬
⎩ ∂x ⎭ ⎩ ∂y ⎭
+ + a h cv ( T s
−T a)
∂x ∂y
In this equation ρa is the air density, Cpa is air heat capacity, uya is air inlet speed, and Ta is air
temperature at any point, K is air thermal conductivity, a is the convection area factor
between the clinker and air, Ts is solid temperature at any point in the cooler. In Equation
(4.31) the left hand side terms represents the net energy input by the air. First two terms in the
right hand side represent the conduction between the air layers and the final term is due to the
convection between solids and air. Since the thermal conductivity of the air is very less and
101
the neighbouring cells are at approximately same temperature so the conductive heat transfer
among the air layers will be negligible.
In Equation (4.31) hcv is convective heat transfer coefficient between solid clinkers and air.
Developing accurate models for convective heat transfer coefficient between solids and air is
most important because the dominating mode of heat transfer in the cooler is convection. The
calculations are based on empirically determined heat transfer coefficient for forced
convection in packed bed given by (Nsofor and Adebiyi, 2001)
Nu = 8.74 + 9.34[6(1 - ε )]0.2 Re 0.2 Pr 0.33 (30 < Nu < 60; 50<Re<120) (4.32)
It is important to note that the Reynolds numbers for commercial clinker cooler are
significantly higher (Re ~ 1000 to 2000) as compared to experimental conditions of Nsofor
and Adebiyi (2001). However, as discussed earlier, there are no other systematic
experimental studies reported to predict convective heat transfer coefficients in clinker
coolers. The empirical correlation (Equation 4.32) was developed for particle sizes close to
one found in industrial clinker coolers and for wide range of temperature conditions as
observed in clinker coolers. Fortunately, the empirical correlation seems to be weekly
dependent on Reynolds number (Reynolds number is to power 0.2). Therefore possible errors
associated with Equation (4.32) are not expected to change the simulation results
significantly (predicted Nusselt number is ~ 50 to 60). Hence Equation (4.32) was used to
predict gas solid heat transfer in clinker coolers on the present model. All the physical
properties are calculated an average temperature of solids and air as Tf = (Ts + Ta)/2. The
system of algebraic linear equations formulated for above model equations was solved using
tri-diagonal matrix algorithm (TDMA).
The model presented in the previous section was used to evaluate the performance of clinker
cooler. To test the framework of the developed model, the length of cooler was set to 11 m
for all the simulations. The inlet temperatures of air and cement clinker entering the cooler
were taken as 300 K and 1625 K respectively (adapted from the data provided by Vikram
Cements). Simulation results for grate speed of 0.1 m/s are shown in Figure 4.11 and Figure
4.12. Figure 4.11 shows the temperatures of top, bottom and middle layer of solids along the
length of the cooler.
102
1800
Bottom Layer
1600
Middle Layer
1400 Top Layer
1200
Temperature, K
1000
800
600
400
200
0
0 0.2 0.4 0.6 0.8 1
Normalized cooler Length
Figure 4.11: Bed temperature along cooler length
1600
1400 Entry layer

Middle layer
1200 Exit layer
Temperature, K
1000
800
600
400
200
0
0 0.2 0.4 0.6 0.8 1
Normalized height of cooler
Figure 4.12: Air temperature along cooler height
Figure 4.12 shows the air temperature in the vertical cells at 3 different position viz. near the
solids entrance, in the intermediate layer and at the solid exit. As expected, due to a higher
103
temperature gradient near the solid entrance the air temperature in the initial layers near the
solid entrance have higher temperatures. The cooling air is supplied to the clinker cooler in 3
sections at different flow rates. The rate of heat transfer in each section, which is a function
of the air flow rate, is different in all the three sections. Therefore the slope of the
temperature curve changes at the location where the sections have common intersection.
Since the rate of heat transfer is highest in the bottom layer the change of slope is prominent
there. Overall, the simulated results seem to be qualitatively reasonable when compared with
the published results (Locher, 2002).
The individual models for pre-heater, calciner, kiln and cooler described in the previous
section have shown reasonable trends when compared with industrial observations. The
individual models were also validated wherever possible. On obtaining a reasonable
agreement and confidence, the models for pre-heater, calciner, kiln and cooler were coupled
with each other to develop a simulator for the entire system.
4.4 Integrated Model

The individual models were coupled with each other to develop a simulator for the entire
system. The schematic of the simulator is shown in Figure 4.13. The required inputs to the
simulator are flow rates and composition of (a) raw meal entering the pre-heater (b) air
entering the cooler (c) coal entering the calciner and the kiln and (d) the material properties
and operating parameters of the individual equipments (for example, kiln RPM, grate speed
of cooler).
Flue gases Hot gases to preheat

Raw meal
P
T.A
Raw Meal
C K.A S.A
Preheated raw
meal K Cooled
Clinker
Partially calcined
raw meal R
Hot Clinker
Inlet Air
Figure 4.13: Schematic of the simulator
104
However, to solve the integrated simulator, it is necessary to know the inlet conditions for the
calciner (flow rate, mass fractions and temperature of solids and air from pre-heater, kiln and
cooler), pre-heater (flow rate and temperature of air from calciner), kiln (flow rate, mass
fractions and temperature of secondary air from cooler and partially calcined raw meal from
the calciner) and cooler (flow rate and temperature of solids from kiln).
To generate these inputs a pre-processor was developed. The function of pre-processor was
two fold. The pre-processor was used to develop good initial guess for the simulator and also
to check for any inconsistency of input data. The preprocessor generated the initial guess (for
mass flow-rates, composition and temperatures of raw meal and air) for the individual models
based on overall material and energy balances. Following parameters were provided to the
preprocessor to achieve this:
1. Percentage calcination occurring in the calciner (P).
2. Temperature of secondary air (Tg,S) and Tertiary air (Tg,T) leaving the cooler.
3. Temperature of air leaving the kiln (Tg, K).
4. Temperature of air exiting the pre-heater to the atmosphere (Tg, P).
5. Temperature of solids exiting the cooler (Ts, R).
6. Heat losses (HLoss, K ) and heat of clinkerization reaction in the kiln (HR, K).
These values are usually known or can be easily available for any cement plant and can
therefore be used to generate good initial guess for faster convergence of solution. The
preprocessor solves mass and energy balance equations as discussed in the following. Based
on the percentage calcination in the calciner, the mass of CO2 produced in calciner was
calculated as
MwCO 2
mCO 2,C = mCaCO 3,i ⋅ P ⋅ (4.33)
MwCaCO 3
where, mCaCO3,i is the total amount of CaCO3 in the inlet raw meal. The mass flow rate of
solids entering the kiln was calculated as
M s , C = M s , P − mCO 2, C (4.34)
105
where, Ms,C is the mass flow rate of raw meal leaving the calciner or entering the kiln, Ms, P is
the mass flow rate of the solids entering the pre-heater. The corresponding mass fraction of
solids species leaving the calciner or entering kiln were calculated as
mCaCO 3,i − mCaCO 3,i ⋅ P (mCO 2,C ) ⋅ ( MwCaO )
xCaCO 3,C = xCaO ,C =
M s ,C ( M s ,C ) ⋅ ( MwCO 2 )
m SiO 2,i m Al 2O 3,i m Fe 2O 3,i

x SiO 2, C = x Al 2O 3, C = x Fe 2O 3, C = (4.35)
M s ,C M s ,C M s ,C
where x is the mass fraction of the component in the raw meal. The amount of clinker leaving
the kiln or entering the cooler Ms,K was calculated as
( M s ,C ) ⋅ xCaCO 3,C
M s , K = M s ,C − ⋅ MwCO 2 (4.36)
MWCaCO 3
Based on overall material balance on kiln, the amount of air leaving the kiln was calculated
as
MwCO 2 MwCO 2
M g , K = M g , S + M s ,C ⋅ xCaCO 3,C ⋅ + M c ,K ⋅ y c,K ⋅ (4.37)
MwCaCO 3 MwCaCO 3
where Mg,S is the mass of secondary air entering the kiln, Mg, K is the air leaving the kiln or
entering the calciner, Mc, K is the amount of coal entering the kiln and yc,k is the mass fraction
of char entering the kiln. The amount of air leaving the pre-heater assembly was calculated as
MwCO 2
M g , P = M g , K + M g ,T + mCO 2,C + M c ,c ⋅ y c ,c ⋅ (4.38)
MwCaCO 3
where Mg,P is the mass of air entering the pre-heater, Mg,T is the mass tertiary air entering the
calciner, mCO2, c is the CO2 produced in calciner due to calcination reaction and Mc, c is the
amount of coal entering the calciner and yc,c is the mass fraction of char entering the calciner.
The temperature of solids leaving the kiln was calculated as
(Ms, R ⋅ C p, s ⋅ Ts, R + M g, T ⋅ C p, g ⋅ Tg,T + M g, S ⋅ C p, g ⋅ Tg,S ) - (Mg, in ⋅ C p, g ⋅ Tg, in )

Ts , K =
(Ms, K ⋅ C p, s )
(4.39)
106
In the above Equation Mg,in and Tg,in, is the mass flow rate and temperature of air entering the
cooler and Ts, R is the temperature of solids exiting the cooler. The temperate of solids entering
the kiln or exiting calciner (Ts,C) was calculated as
(M s, K ⋅ C p, s ⋅ Ts, K + M g, K ⋅ C p, g ⋅ Tg, K + H R, K + H Loss, K - M g,S ⋅ C p,g ⋅ Tg, S - H c,K )

Ts ,C =
(M s,C ⋅ C p, s )
(4.40)
In the above equation, HC,K is heat released due to coal combustion in kiln, HR,K is heat
required for clinker reactions and HLoss, K is the loss from the kiln. The temperature of solids
entering the kiln is essentially same as temperature of gases leaving the calciner (Tg, C).
Finally, the temperature of solids entering the calciner or leaving the pre-heater assembly
(Ts, P) was calculated as
(M s, C ⋅ C p, s ⋅ Ts, C + M g, C ⋅ C p, g ⋅ Tg, C + H calc - M g,K ⋅ C p, g ⋅ Tg, K - M g,T ⋅ C p, g ⋅ Tg,T − H C,C )
Ts , P =
(M s, P ⋅ C p, s )
(4.41)
In the above equation, HC, C is heat released due to coal combustion in the calciner and Hcalc is
the heat required by calcinations reaction. This was easily calculated based on percentage
calcination occurring in the calciner. The heat losses in calciner are negligible as compared to
total heat supplied to the calciner (<5 % of total energy input) and therefore was not
considered in pre-processor calculations. In this way the input conditions (mass, mass
fractions and temperature) for pre-heater, calciner, kiln and cooler were calculated using
preprocessor. The values calculated by preprocessor were passed as input conditions to the
individual models which were then solved iteratively as shown in Figure 4.14.
The iterations were continued till the temperature of solids and gases at exit of individual
components were within error of ±1 %. Suitable under relaxation parameters were used.
Typically 10-20 iterations were required for solution to converge. The convergence results
for two values of under-relaxation parameter are shown in Figure 4.15. An easy to use,
graphical user interface (GUI) based software called RoCKS (Rotary Cement Kiln Simulator)
was developed based on the integrated modules of pre-heater, calciner, kiln and cooler which
is discussed in the nest section
107
User Input Consistence Checks
Dimensions, MOC, Call Preprocessor &
Mass Flow Rate,
Mass Fractions, N Y Generate initial guess
Temperature
Call Sub-models
Update variables
No
Converged
Yes
Post Processing
Figure 4.14: Solution Methodology of the simulator

Calciner exit air temperature, K
1032 1390
Tertiary Air Temperature, K
1380
1031.5
1370
1031 u=0.5
u=1.0 u = 0.5
1360
u=1
1030.5
1350
1030 1340
0 20 40 60 80 100 0 20 40 60 80
Iterationnumber Iteration Number
Figure 4.15a: Calciner exit air temperature Figure 4.15b: Tertiary air temperature
Figure 4.15: Influence of under-relaxation parameter on convergence behavior

(simulations of Vikram cement line I)
108
4.5 RoCKS (Rotary Cement Kiln Simulator)
The integrated modules of pre-heater, calciner, kiln and cooler were included in RoCKS
(Rotary Cement Kiln Simulator) software. The FORTRAN programs implementing the
solution of integrated simulator were modified to develop dynamically linked libraries
(DLL). Two DLL’s, one for the preprocessor and second for the main simulator were
developed. GUI was developed using Visual Basic (VB) to facilitate use of the models
developed in this work. Typical screen shots of RoCKS are shown in figures below.
Figure 4.16: Snap shot of Rotary Cement Kiln Simulator (RoCKS) Software.
109
Figure 4.18: Snap shots of internal windows of (RoCKS)
110
Figure 4.19: Snap shots of Internal windows of (RoCKS)
111
4.6 Simulation Results
The integrated model (RoCKS) presented in the previous section was used to simulate
performances of pre-heater, calciner, rotary kiln and cooler in clinker manufacturing. Based
on the available data on rotary kilns (as discussed in Chapter 3) and available information
from some of the cement industries, a typical clinker manufacturing configuration was
selected as a base case. Some assumptions were made to fill in the gaps in the available data.
The details of selected configuration are given in Tables 2a-2c. Though the developed
mathematical framework is general enough to accommodate temperature dependent physical
properties like heat capacity, at this stage, these properties were treated as constants. The
physical properties of solids and air used in this work are specified in Table 3a. Our prior
simulations of kiln and calciner indicated that the errors in overall energy consumption
associated with the assumption of temperature independent values of specific heat were
within 1%. The operation of the base case (described in Tables 2 to 4) was computationally
studied to understand the various processes occurring in individual units in clinker formation.
On obtaining satisfactory results from the base case, several numerical experiments were
performed using the model for understanding interactions among different processes and for
possible optimization of clinker manufacturing process.
Table 4.4: The Dimensions of pre-heater unit
PREHEATER UNIT
S/No Description Units Values
1 No of pre-Heaters 4
2 Height of cylindrical section m 5
3 Height of conical section m 3
4 Diameter of Cyclone m 3
5 Diameter of cone tip m 1
6 Refractory thickness m 0.13
7 Shell thickness m 0.03
8 Inlet duct height m 1
9 Inlet duct width m 1
10 Diameter of outlet pipe m 1
112
Table 4.5: The Dimensions of kiln
KILN
1 Length m 50
2 Inner Diameter m 3.4
3 Coating Thickness m 0.136
4 Refractory Thickness m 0.2
5 Shell thickness m 0.025
Table 4.7: The physical properties of solids and air
Description Air Raw meal Coal
Thermal Conductivity, W/m K 0.116 0.5 0.5

Emmisivity 0.4 0.9 0.8
Heat capacity, J/kg K 1000 1000 1000
Viscoscity, kg /m s 1e-05 - -
Density, kg/m3 1.3 1500 1000
Char calorific value, J/kg - - 5600
Volatile calorific value, J/kg - - 11900
113
Table 4.6: The Dimensions of cooler
COOLER
1 Length m 11
2 Width m 1
4.6.1 Base case simulation
The predicted results from the simulation of the base case are summarized below. The mass
fractions and temperatures of solids and air in pre-heaters, calciner, kiln and cooler obtained
from the simulation are plotted in Figure 4.20a and Figure 4.20b respectively. It is important
to note that the flow of air is counter current with respect to the flow of solids in the system.
The abscissas of Figure 4.20a and Figure 4.20b denote particular equipment in clinker
formation as discussed below. Abscissa 1 to 4 corresponds to pre-heater assembly. Abscissas
4 to 5 denote the calciner in the system. Abscissas 5 to 15 denote the rotary kiln and 15-18
denote the cooler section. Figure 4.20a shows a plot of mass fractions in pre-hater, calciner,
kiln and cooler (only CaO, C2S and CO2 mass fractions are plotted for the sake of brevity).
Since there is no reaction occurring in pre-heater section, the composition of CaO and CO2 in
this section do not vary. However, in the pre-heater section the raw meal gets heated from
300 K to 1069 K and hot gases from calciner get cooled ( from 1224 K to 539 K) as can be
seen from Figure 4.20b.
As the raw meal passes through the calciner, it gets partially calcined. Therefore CaO
concentration increases in the calciner section as can be seen from Figure 4.20a. Similarly
since CO2 is formed due to calcination and coal combustion, the mass fraction of CO2
increases in the calciner. Coal combustion in the calciner accounts for rise in temperature of
both solids and gas in the calciner (see Figure 4.20b). Remaining clinkerization reactions
occur in kiln. The mass fraction and temperature profiles obtained in kiln (as shown in Figure
4.20a and Figure 4.20b) are similar to previously published results (See Chapter 3;
Mastorakos et al., 1999). Since there is no reaction occurring in the cooler, mass fraction of
solids in the clinker cooler do not vary. However air entering the cooler gets preheated (from
300 K to 1200 K) and solids get cooled (from 1632 K to 476 K) in the cooler section.
114
0.7
0.6
CaO -mass fraction C2S-mass fraction
CO2-mass fraction
Mass fractions 0.5
0.4
0.3
0.2
Calciner
Cooler
0.1 Pre-Heaters Kiln
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Divisons accross the cement clinker process
Figure 4.20a: Mass fraction in pre-heaters, calciner, kiln and cooler in a cement clinker
process.
2000 2000
Solid temperature, K
1500 1500
Gas temperature, K
Gases entering 3rd preheater
1000 1000
Solids leaving 3rd preheater
500 Coole r 500

Calcine r
Pre -He ate rs Kiln
Solid temperature
Gas temperature
0 0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Divisons accross the cement clinker process
Figure 4.20b: Temperature profile across pre-heaters, calciner, kiln and cooler in a
cement clinker process.
115
The predicted energy requirements of individual processes like clinkerization reactions,
losses, melting predicted by the model are listed in Table 4.8. The obtained results are
qualitatively similar to previously published results (Engin and Ari, 2005). The performance
of the overall system was characterized in terms of net energy consumption per unit weight of
product (clinker coming out of the kiln). The net energy consumption (NEC) was calculated
as:
NEC = (ERXN, C + ERXN, K + EMELT, K ) + ELOSS + (EG,OUT + ES,OUT - EG,IN - ES,IN) (4.42)
In the above equation, ERXN denotes the energy required for clinkertization reactions and
subscripts c and k denotes the calciner and the kiln respectively. The term EMELT, K denotes
the energy required for melting in the kiln. ELOSS denotes the summation of energy losses
from pre-heater assembly, calciner and kiln. The other terms denote energy flow rates
(subscripts IN or OUT) for the gas and solid streams (subscripts G or S) which denote the
energy required to raise the sensible heat of the solids. Based on above calculations, the net
energy consumption predicted by the integrated simulator, for these operating conditions was
2635 kJ/ kg clinker (630 kcal/ kg clinker) which seems to be reasonable when compared with
industrial observations. Overall the integrated simulator was able to predict the clinker
manufacturing process in cement industry reasonably well.
116
Table 4.8: Complete energy balance of the system
S/No Description Pre-heater Calciner Kiln Cooler
1 Solid inlet temperature, K 300 1069.2 1214.8 1622.4

2 Mass flow rate , kg 50 50 37.74 32.3
3 Air inlet temperature, K 1214.8 1114.5 1229.9 300
4 Air flow rate, kg/s 60.8 46.7 16.2 45
5 Coal flow rate, kg/s - 2.15 0.9 -
6 Coal inlet temperature, K - 350 350 -
7 Heat with solids in, kJ/kg clinker 463.0 1650.2 1415.3 1622.4
8 Heat with air in, kJ/kg clinker 2297.9 1603.6 615.0 416.7
9 Heat with coal in, kJ/kg clinker - 23.2 9.7 -
10 Combustion of coal, kJ/kg clinker - 1876.7 747.1 -
11 Heat of reaction, kJ/ kg clinker - 1384.5 219.0 -
12 Heat of melting, kJ/ kg clinker - 44.2 - -
13 Heat of solids leaving, kJ/ kg 1650.2 1415.3 1622.4 463.0
clinker
14 Heat of air leaving, kJ/ kg clinker 1014.4 2297.9 1603.6 1415.3
15 Heat of vent air in cooler, kJ/ kg - - - 109.4
clinker
16 Heat with ash, kJ/ kg clinker - 3.22 2.24 -
17 Losses, kJ/ kg clinker 98.9 43.5 140.7 -
117
4.6.2. Influence of key design and operating parameters on net energy consumption
On obtaining a reasonable agreement, the model was used to explore space of design and
operating parameters to understand influence of these parameters on the performance of
clinker manufacturing. All these simulations were carried for a fixed product composition
(C3S mass fraction 0.48 in the product). This was achieved by altering coal flow rate either to
calciner or kiln. This analysis is presented in the section below.
Effect of number of pre-heaters
The effect of changing number of pre-heaters in pre-heater assembly (from 3 to 5) on net

energy consumption was studied. For this simulation the coal in the kiln was adjusted to get
same product composition at the kiln exit. The results for this simulation are shown in Figure
4.21. It can be seen from Figure 4.21 that as number of pre-heaters in pre-heater assembly
increases, solids get pre-heated to a higher temperature before they enter the calciner (see
secondary axis in Figure 4.21). Therefore coal requirement for a fixed product composition
decreases. Thus the net energy consumption decreases as number of pre-heaters increases.
However, the overall capital cost increases by increasing number of pre-heaters in the system.
The developed model will be useful to carry out cost to benefit analysis for introducing
additional pre-heater in the pre-heater assembly.
645 1100
Solid temperature entering calciner, K
Energy consumption, kcal/kg clinker
640
1080
635
1060
630
1040
625
1020
620
615 1000
2 3 4 5 6
Number of pre-heaters
Figure 4.21: Effect of pre-heater number on overall energy consumption
118
Effect of percentage calcination in the calciner
The pre-calcination of raw meal in calciner is an important process in cement process. The
effect of percentage calcination in calciner on net energy consumption was studied. The
percentage of calcination in pre-calciner was varied from 50 to 80. To achieve this, the coal
feed rate to the calciner and the kiln was altered to get the same clinker composition. The
simulation results are shown in Figure 4.22. The net energy consumption is found to decrease
till 70 % calcination and then it increases with further increase in percentage of calcination.
The secondary axis of Figure 4.22 shows that the kiln exit gas temperature also shows a
similar trend. Table 4.9 shows a complete comparison of heat of reaction occurring in kiln
and calciner in this process. The heat of reaction in kiln decreases as the percentage
calcination increases in calciner. The total heat of reaction in kiln is summation of heat of
calcination (endothermic reaction) and heat of clinker formation (exothermic reactions).
When calcination occurs pre-dominantly in the pre-calciner (> 70%), the energy requirements
for reactions in kiln reduce drastically. This causes increase in kiln flue gas temperature and
increase in losses from kiln shell. Therefore the net energy consumption and kiln flue gas
temperature increases if more than 70 % calcination occurs in the calciner.
It is also important to note that the variation in net energy consumption appears to be small in
nature (< 1%) from results shown in Figure 4.22. However, considering the production
capacities of the rotary cement kilns, the overall impact is quite significant. Typically for the
cement kiln under study (with production rate of ~ 3000 TPD) within this calcination range, it
can be easily shown there is an possibility of energy savings of ~ 1.12 MW. The model and
the RoCKS software were thus able to provide valuable clues for determining the optimum
percentage calcinations desired for minimizing net energy consumption.
Table 4.9: Heat of reaction in calciner and kiln
S/No Heat of reaction

1 CALCINATION, % 50 60 70 80
2 Heat of reaction in calciner, kJ/kg 1038.5 1176.6 1384 1618.5
clinker
3 Heat of reaction in kiln, kJ/kg clinker 582.9 433.28 219 -20.28
119
638 1300
Energy consumption, kcal/kg clinker

637
Kiln exit gas temperature, K

1250
636
635
1200
634
633
1150
632
631
1100
630
629 1050
30 40 50 60 70 80 90
% Calcination
Figure 4.22: Effect of percentage calcination on overall energy consumption
Effect of kiln RPM, kiln tilt and grate speed of clinker cooler
The effect of kiln rotational speed and kiln tilt on the overall performance is shown in Figure
4.23a and 4.23b. For these simulations the coal flow rate to the kiln was varied to maintain
constant product composition. It can be seen from Figure 4.23a that as the kiln RPM
decreases, the net energy consumption decreases. Changes in kiln RPM changes the bed
height and the residence time of solids in the kiln as can be seen from Figure 4.23a and 4.23b
(2002.4 s for 3 rpm; 1058.2 s for 5.5 RPM and 703.4 s for 8 rpm). Our simulation results
indicate that it seems to be beneficial to operate kilns at lower RPM as long as adequate
mixing of solids is occurring. From Figure 4.23b it can be seen that energy consumptions in
kilns operated at lower tilt is less as compared to kilns at higher tilt. The grate of clinker
cooler is the important parameter that controls the residence time of solids and subsequently
the heat exchange between hot solids and counter current air in the cooler. We have studied
the influence of varying grate speed on overall energy consumption. The results for these
simulations are shown in Figure 4.23c. It can be seen from Figure 4.23 that the net energy
consumption increases with increasing grate speed. The increase in grate speed reduces
residence time of solids in the cooler. This results in less convective heat transfer between
solids and air as clearly indicated by temperature of secondary air plotted in Figure 4.23.
Therefore simulation results indicate that it is better to operate grates in the cooler at lower
120
speed. The simulations presented here provide useful trends of energy consumption as a
function of key operating parameters in cement clinker process. This results also gives us a
scope to understand the importance of design parameters (kiln tilt) on plant performance and
can be very useful to plant engineers.
640 2500
Residence time of solids in kiln, s

Energy Consumption, kcal/ kg clinker
635
2000
630
1500
625
1000
620
615 500
2 3 4 5 6 7 8 9
Kiln Rpm
Figure 4.23a: Effect of kiln RPM, on overall energy consumption
640 2000
Energy consumption, kcal/ kg clinker
635
Residence time of solids, s
1600
630
1200
625
800
620
615 400
0 0.5 1 1.5 2 2.5 3 3.5
Kiln tilt, Degree
Figure 4.23b: Effect of kiln tilt, on overall energy consumption
121
650 1350
Energy consumption, kcal/ kg of clinker

645 1300
Secondary air temperature, K

640 1250
635 1200
630 1150
625 1100
620 1050
0.06 0.08 0.1 0.12 0.14 0.16 0.18
Grate speed, m/s
Figure 4.23c: Effect of cooler grate speed, on overall energy consumption
Effect of solid loading
The predicted results in the form of net energy consumption and corresponding overall losses
for different solids flow rates are shown in Figure 4.24. It can be seen that the net energy
consumption per unit weight of product decreases as solids flow rate increases. The mass flux
of the product through the kiln also showed similar trend (Corresponding figures for mass
flow rate of 44 kg/s, 47.5 kg/s and 50 kg/s in terms of mass flux are 4.84 kg/m2s, 5.23 kg/m2s
and 5.5 kg/m2s respectively). This is because the net energy loss from the entire system
decreases as the solid flow rate increases (see Figure 4.24). Thus it is beneficial from the
point of view of energy consumption to operate the units with higher solids flow rate. Other
operational concerns like increase in dusting and mixing however need to be considered
while identifying maximum solids flow rate specifically for cement kilns.
122
650 350
Total lose in clinker process, kJ/kg clinker

Energy cons umption, kcal/kg clinker
640 330
630 310
620 290
610 270
600 250
44 45 46 47 48 49 50 51
Raw meal flow rate, kg/s
Figure 4.24: Effect of raw meal flow rate, on overall energy consumption
Effect of coal composition
The effect of varying coal composition to the kiln on net energy consumption is shown in
Figure 4.25. From Figure 4.25, it can be seen that the overall energy consumption does not
change significantly with changing coal composition (ash content 9 %, 15 % and 40 %). For
these simulations the coal flow rate to the kiln was adjusted so that same amount of energy is
supplied to the kiln. Therefore the insignificant change in overall energy consumption does
not seem to be surprising. However, as the coal composition changes, the flame
characteristics in the kiln vary. The predicted flame length by the simulator for varying coal
composition is shown in Figure 4.25. The flame length was calculated by tracking the region
in freeboard where char and volatiles composition in coal go to zero. It can be seen that coal
with higher ash content tend to have a longer flame as compared to coal with lower ash
content. The flame length is a complicated function of amount of oxygen, amount of char and
temperature of gas and particle in the freeboard region. Coals with higher ash content tend
consume oxygen at a slower rate and therefore result in longer flames. Such simulations can
therefore provide useful information to kiln operators to predict the flame characteristics for
wide variety of coal available in the market.
123
650 0.56
Energy comsumption, kcal/kg of clinker

640
Flame length, dimensionless

0.52
630
0.48
620
0.44
610
600 0.4
30 40 50 60 70
Char Percentage, %
Figure 4.25: Effect of coal composition, on overall energy consumption
Effect of secondary shell
Heat losses to the surrounding from the kiln shell by radiation and convection are a
significant source of energy loss in cement kilns and therefore the overall process. These
losses can be reduced by using a secondary shell. The idea is to cover the kiln shell with
another metallic shell having low surface emissivity and thermal conductivity (Engin and Ari,
2005). However, merely covering kiln shell with metallic shell and insulating it can lead to
enormously high shell temperatures. Hence a practical approach to use secondary shell would
be to feed air through the interstitial space of shell and secondary shell to recover the energy
and still operate kilns under realistic conditions (as discussed in Chapter 3). The developed
RoCKS frame work was used to explore the possibility of using such a secondary shell. The
losses in kiln reduced from 140 kJ/ kg of clinker to 1.4kJ/ kg of clinker on applying a
secondary shell and insulation of dimensions and operating conditions specified in Chapter 3.
The net energy consumption reduces from ~ 2635 kJ/ kg clinker to 2493 kJ/ kg clinker (i.e.
630 kcal/ kg clinker to 596 kcal/kg clinker) by using secondary shell and passing air of about
30 kg/s through the interstitial space. If the air coming out of annular space at ~ 496 K can be
utilized within the cement plant (refrigeration, drying of fly ash and so on), the use of
124
secondary shell appears to be promising for reducing net energy consumption in the clinker
manufacturing process.
600 1120
Energy comsumption, kcal/kg of clinker

1115
598
Kiln flue gas temperature, K

1110
596
1105
594
1100
592
1095
590 1090
15 20 25 30 35 40 45
Air flow rate through secondary shell, kg/s
Figure 4.26: Effect of secondary shell, on overall energy consumption
4.6 Conclusions
A comprehensive model was developed to simulate complex processes occurring in pre-
heater, calciner, kiln and cooler for clinker formation in cement industry. The models for pre-
heater and calciner were developed assuming solids and gas to be completely back mixed.
The computational model for the kiln was developed assuming gas and solids as plug flow.
The integrated simulator was converted into simple to use GUI based software for cement
industry, named as Rotary Cement Kiln Simulator (RoCKS). RoCKS was used to simulate
performance of pre-heater, calciner, kiln and cooler for clinker formation. Detailed validation
was unfortunately not possible since adequate industrial data could not be obtained. However,
the model predictions agreed reasonably with industrial observations. RoCKS was used to
understand influence of various design and operating parameters on overall performance.
Specific conclusions based on this computational study are:
• Including an additional pre-heater reduces net energy consumption. The developed
model can be used to evaluate relative benefits of energy savings by additional pre-
heater and required additional capital expenses.
125
• There is an optimum value for percentage of calcination carried out in calciner with
respect to overall energy consumption in clinker manufacture. With the parameters
selected in this work, this optimum value of percentage calcination in calciner is
about 70.
• The simulation results indicated that operating kiln with higher solid loading, lower
RPM, lower tilt and lower grate speed reduces energy consumption per unit
production. The upper limit on solid loading (bed height) and lower limits on RPM
and tilt (mixing and heat transfer) need to be identified based on other practical
issues.
• The use of secondary shell appears to be a promising method to reduce overall
energy consumption, if the hot air generates in such secondary shell (~ 200 0 C) can
be utilized in some other processes in cement plants.
The model was also able to predict kiln characteristics like maximum flame temperature and
overall flame length for coals with different compositions. The models and results presented
here will help in developing a better understanding of clinker manufacturing process and may
provide clues for possible optimization.
126
5. Computational Fluid Dynamics based Models for Transverse Section in
Rotating Kilns
Abstract
This chapter reports development of CFD models to simulate solids motion and heat transfer
in the transverse plane of a rotating kiln. The computational models for flow of solids were
developed by following two approaches. Initially a CFD model for simulating gas-solid
motion in a transverse plane was attempted based on the Eulerian-Eulerian approach. A
pseudo two-dimensional solution domain was considered. Sensitivity of predicted results with
respect to various physical parameters like particle size, restitution coefficient, maximum
packing limit, frictional viscosity was studied. Emphasis was given to predict the shape of
solids in the transverse plane and solids velocities at the surface. Simulated results were
compared with the reported experimental data. As an alternative approach, solid motion and
heat transfer in rotary cylinder was simulated assuming solids as pseudo-homogeneous fluid.
Based on this approach, the computational models were developed to study heat transfer in
partially filled rotating cylinders. The numerical simulations were carried out to understand
the influence of key operating variables like different solid fills, rotational speeds, source
temperatures and bed properties on the temperature gradients in the bed. The models and
presented results will be useful to develop more comprehensive, three-dimensional CFD
models of rotary cement kilns.
Publications based on this work

• Vivek V. Ranade and Kaustubh S. Mujumdar, CFD simulations of solid motion in the transverse plane of
rotating kilns, Third International Conference on CFD in the Minerals and Process Industries, CSIRO,
Melbourne, Australia, 10-12 December 2003.
• Kaustubh S. Mujumdar and Vivek V. Ranade, CFD simulations of heat transfer in transverse plane of
rotating kilns, CHEMCON- Bhubhaneshwar, 2003.
127
5.1 Introduction
The previous chapters reported comprehensive reaction engineering based model for clinker
manufacture in cement industry. The reaction engineering based model for rotary cement
kilns was developed to gain an overall understanding of the kiln behavior. The
phenomenology based kiln model was integrated with computational models for pre-heater,
calciner and cooler model to develop a simulator for cement industry, which was named as
RoCKS (Rotary Cement Kiln Simulator). The RoCKS was able to capture influence of key
design (kiln dimensions) and operating parameters (RPM, solids flow rate, composition, coal
flow rate and so on) on overall performance of cement kilns (energy consumption per unit
weight of product, production rate per unit volume of kiln and so on). Unfortunately, RoCKS
will not able to capture influence of burner design and key operating parameters like ratio of
swirl or axial air etc. on performance of cement kilns. Insight into coal combustion and its
influence on temperature profiles within the kiln required to be known for variety of coal
available in market. In order to ensure the quality of product without jeopardizing the energy
efficiency, it is essential to have a tool, which provides an insight and quantitative guidelines
for manipulating burner operations. Development of a comprehensive CFD framework which
can provide such information for cement kilns was therefore undertaken in this work.
In recent years, CFD based models are being applied to simulate rotary kilns (Mastarokas et
al., 1999; Karki et al., 2000). It is possible to simulate coal burner and free board region of
rotary kiln fairly accurately using the state of the art CFD models (Mastorakos et al., 1999;
Karki et al., 2000). Before we start the development of a comprehensive, three-dimensional
CFD model of a rotary kiln, it was thought desirable to understand flow, mixing and heat
transfer in a transverse plane of a rotary kiln. One of the objectives of this exercise was to
determine and to evolve a rational basis for possible simplifications which may be used with
the complete 3D models. The computational models for flow of solids in a transverse plane
were developed by following two approaches. Initially flow models using the Eulerian-
Eulerian approach were attempted to model solids motion in a transverse section. Sensitivity
of various physical parameters like particle size, restitution coefficient, maximum packing
limit, frictional viscosity on flow of solids was studied. The predicted results for velocities
were compared with available experimental data. The solid motion in transverse section was
also modeled assuming solids as pseudo-homogeneous fluid. The computational model based
on this approach was extended to study heat transfer in rotating cylinders. In the next section
128
computational model and subsequent results for solid motion in rotary cylinders are
presented. This will be followed by discussion on modeling and simulated results for heat
transfer in transverse section of rotary cylinders.
5.2 Motion of solids in transverse section of rotary kilns

Rotating kilns are widely used in the chemical and process industries as mixers, dryers and
reactors. The ability of these equipments to handle a large volume of material along with
broad particle size distribution makes these equipments very popular and common in process
industry. The motion of solids in rotary kilns is extremely complex as the solids
simultaneously move forward due to gravity along with enforced rotation due to revolving
equipment. The solid motion and mixing in the transverse plane determines the surface
renewal rate and hence the rates of bed-freeboard heat and mass transfer, as well as chemical
reactions. This was the motivation to develop computational models for simulating the
motion of solids in transverse section of rotary kilns. Such models can be coupled with
reaction engineering and heat transfer models to develop comprehensive and reliable models
for rotary kilns. Also such models can be used to explore possible simplifications which may
be used while developing a 3D modeling framework.
Earlier attempts of mathematical modeling to predict solids motion in rolling mode may be
classified in three categories [Ding et al., (2001)]. The first category is the development of
simple models based on geometrical and stochastic analyses, and well-mixed tank theory for
example Saeman (1951), Kramers and Crookewit (1952). These models were based on the
assumption that particles in a rolling bed move in a circular motion with the rotating kiln, and
then fall down from the surface of the bed in a thin layer. The time taken to fall down is small
when compared to the time taken by particles to travel up in the bed. The basic model
[Saeman (1951)] predicted experimental data well, and the model was further refined to
predict axial moment of particles with different bed fills, taking into account the time for
particles to fall down the surface of the bed. The second category is discrete particle models
based on Lagrangian approach [Walton and Braun (1993), Cleary et al., (1998)]. The third
category models are based on the Eulerian approach, which assumes solids behave as
continuum and flow properties like solids concentration and velocities are continuous
129
function of position. [Perron an Bui (1990), Ferron and Singh (1991), Boateng and Barr
(1996)].
In rotary kilns the dispersed phase has a high volume fraction, generally greater than 10 %.
Since it will be computationally demanding to use the discrete particle models, the solid
motion was modeled using the Eulerian-Eulerian approach i.e. treating both solids and gas as
continuous phases. In this approach we develop mass, momentum and energy balances and
associated constitutive equations based on volume averaging. The constitutive equations for
granular flows however are not well established. Boateng and Barr (1996) developed a
model, which employed boundary layer approach in deriving governing equations. They
presented their integro-differential set of equations, which contained granular temperature
term. Recently Ding et al. (2001) developed a more sophisticated model considering a quasi-
static contribution, which becomes important at low rotational speed. This term was not
considered in Boateng’s model. The above-mentioned models use a polynomial function for
the velocity profile in the active layer to get the solution to the integro-differential equations.
This velocity profile is mostly obtained from experiments. The models assume surface of
solids to be flat, which is not really the case especially near the walls. Most of the work was
carried out at low to medium rotational speeds. Further, these models are restricted to the
rolling mode of operating regime. More generalized models need to be developed which may
account for different particle sizes, shapes and properties at higher rotational speeds.
Recent advances in understanding of granular flows, numerical techniques and computing

resources may make computational fluid dynamics (CFD) based modeling a better approach
to simulate transverse motion of solids in rotating kilns. Appropriate models to represent
interphase coupling and frictional viscosity are needed. If such models are developed, these
models will not require velocity profile from experiments and could be used to make ‘a
priori’ predictions. These models could also be extended further to understand flow in other
regimes as well. CFD based simulations of flow in a typical transverse section of rotating
kilns, rotating at different speeds carried out initially are shown in Figure 5.1. It can be seen
that the simulations qualitatively captured the different flow regimes in rotating kiln for
varying rpm (The regimes for corresponding rpm is mentioned in legend in Figure 5.1).
130
These initial simulations have given encouragement to develop and to use CFD based models
to predict flow in the transverse section.
Figure 5.1: CFD simulations of different regimes in rotary kilns
(1 rpm: Slipping; 17 rpm: rolling; 55 rpm: Cascading; 100 rpm: Centrifuging)
In this chapter, we report development of a CFD based models to simulate solids flow in a
kiln operating in a rolling mode. Several numerical experiments were carried out to study
effect of various physical parameters like density of solids, particle size, restitution
coefficient, angle of internal friction, maximum packing limit, etc on the flow. Numerical
simulations were compared with the available experimental data. The model and results will
be useful for understanding complex flow in the transverse plane of rotary kilns.
5.2.1 Model description

The Eulerian-Eulerian approach was used for the two phases in which both solid and gases
were treated as continuum.
131
The continuity equation for each phase is given by
∂ (α k ρ k )
+ ∇ ⋅(α k ρ k U k ) =0
∂t
(5.1)
The momentum balance for each phase is written as
∂ (α k ρ kU k )
+ ∇ ⋅(α k ρ kU kU k ) = − α k ∇p − ∇ ⋅ (α kτ k ) + α k ρ k g + Fk + Fg
∂t
(5.2)
Where, αk represents the volume fraction of each phase, ρk is the density of kth phase. Uk is
the velocity of kth phase. P is a mean pressure shared by all the phases present in the system.
There is no inter-phase mass transfer. The mass balance Equation (5.1) therefore, does not
have any source term on the right hand side. Left hand side of Equation (5.2) represents the
rate of change of momentum for kth phase. The right hand side represents pressure forces,
drag, gravitational acceleration, average shear stresses and inter-phase momentum exchange.
The term Fk represents the drag interaction between the two phases. The drag force
interaction between the two phases however was found to have a negligible influence on the
flow of solids. Hence all the simulations were carried out without drag interaction between
the two phases.
The equation describing solids flow, [Equation (5.2)] has a shear stress term in which,
appropriate stress terms needs to be incorporated. Frictional and collisional interactions may
govern this term [Boateng (1998)]. However, in case of rotary kilns, the particles interact
with each other largely through enduring frictional contact between multiple neighbors and to
a lesser extent through collisions. These frictional interactions indeed play a very important
role in dense phase gas-solid flows. Particulate stresses are developed in such dense flows by
frictional interactions between particles at points of sustained contact. It has been shown that
frictional stresses play a critical role in maintaining stable operations of dense solid flows
[Srivastsva and Sunderesan (2002)]. Recently Shrivastava and Sunderesan (2003) have
presented a frictional-kinetic rhoelogical model for dense assemblies of solids in gas solid
flows. They have modeled frictional stress model as proposed by Schaeffer (1987) and
modified it to account for strain rate fluctuations. Schaeffer (1987) assumed that granular
material is non-cohesive and that it follows a rigid-plastic rhoelogical model given by:
132
S
α s τ s = − p F I + A( p F ,α ) (5.3)
S :S
where
σ sf,xx + σ sf,yy + σ sf,zz
pF =
3 (5.4)
1
2
( 1
S = ∇U + ( ∇U )T − ( ∇ ⋅ U )I
3
)
and A is a function to be specified. A variety of models have been described for A(pF , α) in
the soil and granular mechanics literature. This model was modified this frictional-kinetic
model and presented a model for strain rate fluctuations as (Shrivasta and Sunderesan (2003):
S
α s τ s = − p F I + A( p F ,α ) (5.5)
S :S +T / d2
Simulated results obtained with these models are discussed in the following sections.
5.2.2 Geometry, grid and solver
All the numerical simulations were carried out on a cylindrical geometry with a 400 mm
diameter and 3.5 mm in length. Symmetry boundary conditions were specified to the
sidewalls of the cylinder. The mesh for this geometry generated using commercial grid
generation software (GAMBIT of Fluent Inc.) is shown in Figure 5.2.
Figure 5.2: Computational grid used for simulations
133
The grid was refined in the lower section of cylinder where gradients are expected to be
steeper. The simulations were carried out in commercial CFD code FLUENT 6.0.12.
Simulations were carried out with time step of 0.0005 s and carried out until the surface
velocity of the solids reached a steady state.
In the next section we outline results of a sensitivity study of various physical parameters like
density of solids, maximum packing limit, restitution coefficient, particle diameter, angle of
internal friction, frictional viscosity on the solids flow. Numerical simulations were also
carried out for different frictional viscosity models. In all the numerical simulations carried
out emphasis was given to two factors:
1. Shape of the solids in transverse section.

2. Velocity of the solids at the surface. (Tangential velocity at that point)
Ding et al. (2001) have reported experimental data on the flow of solids in the transverse
direction in rotating drums for the rolling mode of operation. A sample of these results is
shown in Figure 5.3. Their experimental data was used to evaluate the CFD model developed
here.
Figure 5.3: Experimental results of Ding et al. (2001)
5.2.3 Numerical experiments and results

The nature of solids flow in rotary kilns can change significantly with particle characteristics.
It is indeed desirable to have tools to make ‘a priori’ estimates of the effects of different key
parameters on the generated flow in a kiln. To quantify the influence of different parameters,
134
detailed sensitivity studies were carried out using the CFD model. For all the numerical
experiments presented here experimental conditions of Ding et al., (2001) were used (1.7
rpm; 12 % solids fill). Initially, effect of particle diameter and restitution coefficient on flow
was studied. However, for a dense bed of solids, as found in a kiln, frictional interactions are
expected to be more important than these parameters. Influence of key parameters governing
the frictional interactions, namely, maximum packing limit of solids (0.6), angle of internal
friction (30o) and frictional viscosity (Equation 5.3) was computationally studied (values in
parenthesis denote base values at which numerical simulations were carried out). The
simulations and their results are presented below.
Effect of particle diameter
Particle diameter sensitivity was studied by considering three sets of particles with diameters
1 mm, 1.5 mm and 2 mm. Other parameters like density, restitution coefficient between
solids maximum packing limit were kept constant. Simulation results are shown in Figure
5.4.
Figure 5.4: Effect of particle diameter on flow on shape
135
From Figure 5.4, it can be clearly seen that effect of change in particle diameter on the shape
of solid bed in transverse section is insignificant. The surface velocities however were found
to be dependent on particle diameter as can be seen from Figure 5.5. It may be noted that the
velocities shown in Figure 5.5 are for the iso-surface of solids volume fraction of 0.5. Small
changes in the shape of the surface and inherent numerical issues involved at the sharp
interface might have caused the differences in the results predicted for these three particle
sizes. The x-axis is the chord length of the free surface of solids (chord AB in Figure 5.3). In
the figure at some points the velocity becomes negative which means that solids have
changed their direction in that region.
0.16
0.14
1 mm
0.12 1.5 mm
2 mm
0.1
0.08
Uo, m/s
0.06
0.04
0.02
0
0 0.05 0.1 0.15 0.2 0.25 0.3
-0.02
-0.04
X, m
Figure 5.5: Effect of particle diameter on surface velocities of solids
Effect of restitution coefficient
Three values of restitution coefficients 0.8, 0.9 and 0.95 were used to study influence of
restitution coefficient on flow of solids. The simulated results are shown in Figure 5.6. It can
be seen that within the considered range the influence of restitution coefficient on shape of
the solid bed in the kiln seems to be more or less same.
136
Figure 5.6: Effect of restitution coefficient on the flow
Effect of maximum packing limit on flow

Maximum packing limit of 0.6 and 0.4 were used to see the effect of the solids packing on the
resulting flow. Results obtained are shown in Figure 5.7.
Figure 5.7: Effect of maximum packing limit on flow
137
From Figure 5.7 it can be seen that although there is not much of effect on the shape of the
solids flow, loosely packed solids tends to move upwards in the rotating kiln.
Effect of angle of internal friction
The numerical experiments were carried out using three values of angle of internal friction 3o,
30o and 60o. Since frictional interactions are expected to be high in flow of solids in rotary
kilns (which will also be discussed in the next section), the effect of angle of friction was
studied on a much wider range. Figure 5.8 shows results of shape of solids obtained for
different angle of internal friction. From Figure 5.8, it can be seen that angle of internal
friction has a significant effect on flow of solids in terms of shape. Similar effect was found
on surface velocities. The effect of angle of internal friction on solids flow is complex ands
needs to be understood in further detail. However, it can be concluded at this point that
friction indeed plays an important role in the flow of the solids in rotating kilns
Figure 5.8: Effect of angle of internal friction of shape
Effect of frictional viscosity
Initially simulations were carried out using constant frictional viscosity values of 10, 25 and
100 Pa-s. The results obtained are presented in Figure 5.9 and Figure 5.10.
138
Figure 5.9: Effect of frictional viscosity on shape.
From Figure 5.9, it can be seen that frictional viscosity indeed significantly affects the shape
of solids in the transverse plane. The shape of solids is widely different at frictional viscosity
of 100 Pa-s than that at frictional viscosity of 10 Pa-s. The predicted surface velocities of
solids were compared with each other in Figure 10. Experimental data of Ding et al. (2001) is
also shown in Figure 5.10 to provide a reference. The physical properties of the solids,
operating conditions used in these simulations were the same as those reported by Ding et al.
(2001). It can be seen from Figure 5.10 that there is a significant difference in the surface
velocities for different frictional viscosities. Although, frictional viscosity in the range of
about 10 Pa-s gives a better agreement with the experimental data as compared to the other
values, around this value of viscosity the shape of solids in the transverse section did not
match with the experimentally reported shape (see Figure 5.3). For frictional viscosity of 100
Pa-s, the shape seems to be satisfactory but the surface velocities did not match with the
reported values. These initial simulations clearly indicates that further work on modeling
frictional viscosity is needed to predict the flow with adequate accuracy.
139
0.15
Ding experimental data
0.13 Ding (theorotical)
Fric-Visc-10
0.11 Fric-Visc-25
Fric-Visc-100
0.09
0.07
Uo (m/s)
0.05
0.03
0.01
-0.010.00 0.05 0.10 0.15 0.20 0.25 0.30
-0.03
-0.05
x (m)
Figure 5.10: Effect of frictional viscosity on surface velocities
Frictional viscosity models
Simulations were carried out using frictional viscosity models proposed by Schaeffer (1987)
and Shrivastava and Sundaresan (2003). The simulation results are shown in Figure 5.11.
Figure 5.11: Effect of frictional viscosity on shape
140
From Figure 5.11 it can be seen that using frictional viscosity proposed by Schaeffer (1987)
and Shrivastava and Sundaresan (2003), the predicted shape of the solids did not match with
the experimentally observed shape. The angle of repose obtained by simulations is about 8o
which further deteriorates as solids move down the surface, where as the experimentally
obtained angle of repose is about 25o.
Comparison of solids velocity at the surface is shown in Figure 5.12. Close examination of
Figures 5.11 and 5.12 indicates that prediction of the shape of solids bed in the kiln and
prediction of surface velocity of solids are intimately linked with each other. It can be seen
that, in the top portion of the solids bed, where angle made by the predicted bed surface is
closer to that observed in the experiments, the predicted velocities are in good agreement
with the experimental data. In the bottom portion, the predicted shape of bed is too flat
compared to that observed in the experiments. Consequently, the predicted velocities are
much smaller than the experimentally observed values. The right prediction of the shape of
the solids bed is crucial for better predictions of solids motion in the kiln.
The velocity profiles obtained in the interior of the bed using Schaeffer viscosity model is
shown in Figure 5.13a. Although the surface velocities could not be captured well by the
described computational models the prediction of interior velocities was not bad.
141
0.13
0.11
0.09
0.07
Uo (m/s)
0.05
0.03
0.01
-0.010.00 0.05 0.10 0.15 0.20 0.25 0.30

Ding experimental data
Ding (theorotical)
-0.03 Schaeffer (1987)
Sunderesan Correction
-0.05
x (m)
Figure 5.12: Comparison of surface velocities for frictional viscosity models
1.2
1
Cfd
0.8 Ding
0.6
0.4
U/Uo
0.2
-0.2
-0.4
-0.6
0 0.2 0.4 0.6 0.8 1 1.2
y/Hx
Figure 5.13a: Comparison of bed interior velocities predicted by Schaeffer’s model
142
Solids as pseudo fluids
The previous section of the chapter discussed about modeling solid motion in transverse
section of rotary kiln using different frictional viscosity models. However, these simulations
were not very successful from the point of quantitative prediction of velocity profiles in the
bed. Computational models were also developed by assuming solids as pseudo-homogeneous
fluid to predict motion of solids in transverse plane. Simulations were carried for the identical
geometry and operating conditions for rotary kiln as described in previous section (Ding et
al., 2001).
1.2
exponent, n =0.5
1
exponent, n = 1
Experimetal values (Ding et al., 2001)
0.8
0.6
u/uo, m/s
0.4
0.2
-0.2
-0.4
-0.6
0 0.05 0.1 0.15 0.2 0.25 0.3
y/Hx
Figure 5.13b: Comparison of bed interior velocities assuming solids as pseudo fluids
For these simulations however, solids were assumed as fluids with constant viscosity and
power law viscosity. The results obtained from these simulations were qualitatively similar to
those obtained in the previous section. The velocities obtained at the bed surface did not
match qualitatively with experimental results. A typical comparison of simulated values of
velocities in the bed interior for one of the case assuming power law viscosity (n = 0.5 and n
= 1 i.e. constant viscosity) is shown in Figure 5.13b. These results seem to be comparable
with results obtained for Schaffer’s frictional viscosity models (see Figure 5.13a). It can also
143
be seen from these simulation that the differences in predictions of velocities considering
power law viscosity and constant viscosity for fluids were insignificant. Considering the
relatively similar performance of model developed in this work (based on the Eulerian-
Eulerian approach) and the pseudo-homogeneous approach, the latter approach was used to
study heat transfer in transverse section of rotary kilns.
5.3 Heat transfer in transverse section of rotary kilns

The heat transfer in rotary kilns is very important as determines its overall performance in
terms of product quality and uniformity. The temperature gradients, if present in radial
direction can directly affect the kiln performance in terms of product uniformity. Hence it is
essential to develop a model which can predict heat transfer (for a set of operating conditions)
in transverse section of rotating kilns. Therefore, a 2D model was developed to study
temperature gradients in transverse section of bed. This section reports the development of
CFD models to simulate heat transfer in the transverse plane of a rotating kiln. In these
simulations solids were treated as pseudo homogeneous fluids as described earlier. The idea
was to check if CFD models are able to predict temperature gradients in bed reasonably well
by assuming solids as pseudo-homogeneous fluid. Numerical simulations were carried out to
see the influence of key operating variables like different solid fills, rotational speeds, source
temperatures and bed properties on the temperature gradients in the transverse plane. A
fictitious heat source was assumed in the central region of the cylinder to provide heat to the
solid bed. Heat transfer occurs by all possible modes i.e. conduction, convection and
radiation. The bed of solids receives heat from the free board gases by radiation and
convection. Along with this, the bed receives heat from kiln walls by radiation and by
conduction. Also, there is convective and radiative heat transfer between the free board gas
and kiln internal walls. One needs take into account all the above mentioned modes of heat
transfer in the model for accurate prediction of temperature gradients in the transverse plane.
5.3.1 Model description
In this work a pseudo-2D model was formulated using the CFD framework to study the heat
transfer. The cylinder geometry and grid used for numerical simulations is explained in the
next section. A fictitious heat source at constant temperature is assumed in the central region
of the cylinder to provide heat to the solids bed. The heat given by this source to the bed is
analogous to the heat given by the flame in the rotary kilns.
144
(a) (b)
(c)
Figure 5.14a: Different solid fill for simulation study
The density of solid bed is assumed to be constant along with material properties. In this
work, we have assumed solids to be pseudo homogeneous fluids with constant viscosity.
Chemical reactions are not considered at this stage. But, the overall framework is such that
they can be incorporated at a later stage. The cylinder outer wall is given rotational boundary
condition and heat loss through outer wall is modeled using convection and radiation. The
sidewalls of cylinder are given symmetric boundary conditions to reduce required
computations. Ambient temperature is assumed to be 300 K and convective heat transfer
145
coefficient is considered to be about 25 W/m2 K. Mass, momentum and energy conservation
equations are given below.
Bed region
Mass conservation equation:
∂ρ
+ ∇ ⋅ ( ρu ) = 0
∂t
Momentum conservation equation:
∂ ( ρu )
+ ∇ ⋅ ( ρuu ) = −∇p − ∇ ⋅ σ + ρg
∂t
Energy conservation equation:
∂T Nc
ρC p + ρC pu ⋅ ∇T = ∇ ⋅ (k∇T ) + ∑ Ri (−ΔH r ,i )
∂t i =1
Free board region

The mass and momentum conservation equations remain same with appropriate changes in
material and their properties. The energy conservation equation with required changes is
given below
Energy conservation equation
δ ( ρE )
+ ∇ ⋅ ( v ( ρE + p )) = ∇ ⋅ ( k eff ∇T − Σh j J j + (τ eff ⋅v )) + S h
δt
The first three terms on the right hand side of the equation represent heat transfer due to
conduction, species diffusion and viscous dissipation respectively. Sh represents the heat
source due to chemical reactions or other volumetric heat sources. Here in this case the
fictitious heat source is the heat source to the solids. Radiation is modeled using the simple P-
1 model. This is the simplest case of more general P-N model, which is based on the
expansion of the radiation intensity into an orthogonal series of spherical harmonics.
A two dimensional geometry for the cylinder is considered in this study. The cylinder interior
contains solid bed with three layers of solids viz. coating, refractory and steel shell (Rin= 3.14
m, Rin, refr= 3.4 m, Rin, sh= 3.8 m, Rout, sh= 3.85 m). Numerical simulations are carried out at
146
three different heights (h ≈ R, h = 2R/3, h = R/3, a, b and c in Figure 5.14a respectively) with
various rotational speeds, source temperatures and properties of bed material. The mesh for
this geometry generated using commercial grid generation software (GAMBIT of Fluent Inc.)
is shown in Figure 5.14b. The simulations are carried out in commercial CFD code FLUENT
6.0.12. Steady state simulations were carried out till the heat flux given by the source and
outer wall (steel shell) of cylinder is same.
Figure 5.14b: Sample of grid used for simulation
5.3.2 Numerical experiments and results

Several numerical experiments were performed to see the influence of important operating
variables like different solid fills, varying rotational speeds, source temperatures and bed
properties on the temperature gradients in the solids bed. In all the numerical experiments, the
source is given a fixed temperature for a particular set of experiments. Initial attention was
focused to analyze the temperature gradients in bed (assumed as pseudo fluid) by variation in
operating parameters like solids fill, rotational speed and source temperature. Radiation is
expected to have a significant effect on the heat transfer. Hence, in the next set of simulations
147
influence of absorption coefficient of the freeboard gas on temperature gradients is studied.
Finally thermal conductivity of solids is changed to see the effect on the heat transfer.
Effect of bed height and rotational speed

The solids fill in rotary kilns has a significant effect on heat transfer in transverse section.
Higher solids fills will tend to have higher temperature gradients and require more energy
requirements. Lower solids fill not only results in lower production capacity but also may
have an impact on life of refractories and kiln shell. We consider in our study three different
solid fills, as shown in Figure 5.13. The numerical simulations are carried out for 2 different
rotational speeds 1 rpm and 5.5 rpm and stationary cylinder as well. The source temperature
is kept at 1700 K. Such high temperatures are often encountered in rotary kilns. The results
for these 9 numerical experiments are shown in Table 5.1.
Table 5.1: Influence of different solid fills

Stationary 1RPM 5.5 RPM
High Height 3 2.4 0.5
Medium Height 2.4 0.6 0.4
Low Height 0.2 0.1 0.3
Figure 5.15 shows contours of temperature in the bed for different solids fill (column 1 of
Table 5.1) when bed is not given any rotation. Table 5.1 gives quantitative estimation of
temperature gradients with in the bed. The gradients in the bed are higher in cases of
stationary bed as there is no mixing of the solids. The gradients reduce when the cylinder is
given rotation as mixing is enhances. The temperature gradients in the solid bed increases as
the height of the bed increases, which is as expected.
148
Figure 5.15: Contour plots of temperature for different solids fill
Effect of source temperature

It can be seen from above analysis that the temperature gradients are quite small (only a
difference of 3 K for even 50 % solids fill). The reason for this could be that the source
temperature being at very high temperature. The temperature in kiln varies along the kiln
length for industrial kilns and hence it is necessary to analyze results at lower source
temperatures as well. Therefore, 2 simulations are carried out in order to see the effect of
source temperature on heat transfer. Three source temperatures considered here are tabulated
below along with the results. For the above runs, the kiln is not given any rotation and
simulations are carried on high solids fill (h ≈ R) is considered as it is expected to give largest
temperature gradients. As can be seen from the table, the temperature gradients rise with
decreasing the source temperature. Also, as the source temperature decreases the overall heat
the bed will receive will decrease. Hence the average temperature of bed will decrease if
source temperature decreases. This is what we get from our simulations as well.
149
Table 5.2: Influence of source temperature
High height (stationary)
Source Temperature ΔTMAX (K)
Temperature Range (K)
1700 K 1666– 1669 3
1350 K 1300 – 1306 6
1000 K 917 – 930. 13
Effect of rotational speed

The effect of rotational speed on the gradients is already mentioned. We can appreciate its
role better in case of lower source temperatures. Two runs are carried out at lower source
temperature for rotational speeds of 1 rpm and 5.5 rpm. The results for these runs are
tabulated below.
Table 5.3: Influence of rotational speed

High height
RPM Temperature ΔTMAX (K)
Range (K)
Stationary 917 – 930 13
1 rpm 924- 927 3
5.5 rpm 924-924.5 0.5
From the results it can be seen that the temperature gradients slowly vanish as the rotational
speed increases. Interesting thing to note is that even at 1 rpm, temperature gradients decrease
significantly as compared to stationary case. This is also found in other simulations. This
probably means, even for small rotation speeds uniform temperature is achieved in the
transverse section of the kiln.
Effect of radiation absorption coefficient

From the literature, it is seen that radiation has a significant contribution to heat transfer in
transverse plane than convection and conduction. The effect of radiation on total heat transfer
150
could be verified by manipulating absorption coefficient of the freeboard gas. The absorption
coefficient of the freeboard gas, which is assumed to be 0.7, is reduced to 0.07 in a couple of
steps. Although this seems to be a bit unrealistic, this will clearly state the roll of radiation in
the heat transfer. The results for these runs are tabulated below. From the above runs, the
dominant role of radiation is displayed for heat transfer in rotary kilns (Compare run 3 in
Table 2 and run 3 in Table 5.4). It can be seen that as absorption coefficient decreases the
heat received by the bed due to radiation decreases. The result of this is the overall bed
temperature reduces significantly (by about 150 K).
Table 5.4: Influence of Radiation

Source Temperature ΔTMAX (K)

Temperature Range (K)
1700 K (absorption 1666– 1669 3
coeff = 0.7)
1700 K (absorption 1565 -1596 31.
coeff = 0.07)
1000 K(absorption 727.6 –826.5 98.9
coeff = 0.07)
Effect of effective thermal conductivity

Finally the effect of the change in clinker properties is considered. The clinker thermal
conductivity is increased from 0.5 to 200 W/m K. This physically would simply mean that
the mixing in bed is very good. The effect of using higher thermal conductivity is thus same
as that of giving rotation to the rotary kiln. The idea behind this is to check if effective
thermal conductivity could be used instead of giving rotational boundary condition to the
shell to reduce the computational requirements. Following is the comparison of the results for
true clinker thermal conductivity and enhanced clinker thermal conductivity with no rotation
given to the kiln.
151
Table 5.6: Effect of effective thermal conductivity
Thermal Temperature ΔTMAX (K)

conductivity of Range (K)
clinker
0.5 W/ m K 1566 - 1597 31
200 W / m K 1609 - 1621 12
The drastic increase in thermal conductivity of clinker has resulted in less temperature
gradient as expected. However, the temperature gradients almost fade out when rotation
boundary condition is specified. Hence a more quantitative relation between the rotational
speed and effective thermal conductivity needs to be developed. The effective thermal
conductivity should also be implemented carefully as the overall bed temperature increases
when effective thermal conductivity is used instead of rotational boundary condition.
5.4 Conclusions
Computational models for modeling motions of solids and heat transfer in transverse section
of rotating cylinders were developed in this study. Following are the conclusions of the
present work.
• A detailed sensitivity study of various physical parameters like diameter of solids,
restitution coefficient, maximum packing limit, angle of internal friction, frictional
viscosity revealed that frictional viscosity has the most significant influence on
flow of solids in transverse plane.
• Solid flow was simulated using a constant frictional viscosity initially and then by
using models proposed by Schaeffer (1987) and Shrivastava and Sundaresan
(2003) with the CFD models. However, the simulation results did not agree with
the experimental results quantitatively. Use of simpler models like pseudo-
homogeneous assumption for the solids phase were found to give results similar to
those with much more complicated models. Further work is needed to accurately
152
model motion of the solids in bed of rotating kilns. In absence of this, all the
subsequent cement kiln simulations were carried out by treating solids as pseudo
homogeneous phase.
• Numerical study carried out considering solids as pseudo fluids predicted
reasonable trends for temperature gradients in bed of rotating kilns at least
qualitatively. Further work is required in order to have quantitative validation in
this respect.
153
6. Computational Fluid Dynamics based model for Rotary Cement Kilns
Abstract
In this chapter, we report a comprehensive computational fluid dynamics (CFD) based model
to capture key transport processes in rotary cement kilns. Separate but coupled
computational models were developed for the bed and the freeboard regions of the rotary
kiln. The complex swirling air flow produced by kiln burners, coal combustion, gas phase
combustion of volatile matter and radiative heat transfer in free board region were modeled.
The clinkerization reactions in the bed region were modeled assuming solids as pseudo
fluids. Coating formation in cement kilns (for both bed and freeboard regions) was
considered. Appropriate source and sink terms were developed to model transfer of CO2 from
the bed to the freeboard region due to calcination reaction in the bed region. The developed
bed and freeboard models were coupled by mass and energy communication through
common interface. These coupled computational models were able to predict the available
data from industrial kilns and previously published results quite satisfactorily. The
computational models were also able to predict the intricacies of burning zone of rotary
cement kilns for changing burner operational parameters like axial to swirl ratio and oxygen
enrichment. The developed approach, computational models and simulation results will not
only help in developing better understanding of cement kilns but also provide quantitative
information about influence of burner design and other design parameters on kiln
performance.
Publication based on this work

• Kaustubh S. Mujumdar and Vivek V. Ranade, “CFD Modeling of rotary cement kilns”
accepted for publication in Asia Pacific Journal of Chemical Engineering
• Vivek V. Ranade and Kaustubh S. Mujumdar, March 2006, “Gain an insight into coal
fired cement kilns using CFD”, The Chemical Engineer, UK
154
6.1 Introduction
This chapter discusses development of comprehensive CFD model for rotary cement kilns.
As discussed earlier, the developed 1D model (in Chapter 3) cannot adequately capture
influence of burner design and key operating parameters like ratio of swirl to axial air,
oxygen enrichment, etc. on flame characteristics and performance of cement kilns. The
knowledge of influence of these parameters on coal combustion, flame characteristics and
temperature profiles within the kiln is essential to ensure optimum performance of kilns.
Computational fluid dynamics (CFD) framework offers promise of becoming such a tool.
Development of a comprehensive CFD framework which can provide an insight and
quantitative guidelines for burner design and optimization for cement kilns was therefore
undertaken in this work. The complexity and co-existence of a wide range of spatio-temporal
scales and processes in rotary cement kilns demand different and unconventional approaches
for developing CFD models. Some attempts of developing CFD models for rotary cement
kilns have been made (for example, Kolyfetis and Markatos, 1996; Karki et al., 2000). These
models were limited only to predict the overall behavior of cement kilns. These models did
not consider some of important processes like coating formation or modeling clinkerization
reactions. This chapter reports the development of CFD models, which adequately accounts
for most of the relevant processes occurring in cement kilns and provides a tool to predict the
influence of burner design and operational parameters on the overall performance.
In the next section of the chapter, we discuss the computational model in detail. Previously
published CFD models were critically reviewed before presenting the approach and model
developed in this work. A solution methodology and simulated results are discussed
thereafter. The computational model was then used to understand influence of various burner
operating parameters on the kiln performance. Key conclusions based on this work are
discussed at the end.
6.2 Computational Model

Cement kilns are complex systems which involve occurrence of several simultaneous
processes in both bed and freeboard region. It is therefore essential to first identify key issues
and use appropriate methodology to develop tractable computational models for rotary
cement kilns. The main key issues which need to be considered while developing
comprehensive model for cement kilns are flow, heat transfer and reactions occurring in bed
155
and freeboard region. These issues are discussed in detail chapter 2 and will not be repeated
in this chapter. It is important to note that most of the previously published CFD models do
not consider these key issues during model development as is discussed below.
6.2.1 CFD models for cement kilns

Due to complexity of physics involved, occurrence of multiple phases with large number of
reactions in bed/freeboard regions, very few CFD models have been published for rotary
cement kilns (Kolyfetis and Markatos, 1996; Mastorakos et al., 1999; Karki et al., 2000).
Most of these computational models do not account the main key issues simultaneously in a
single framework as discussed in Chapter 2. Mastorakos et al. (1999) developed a CFD based
model for cement kilns, which included combustion, radiative heat transfer, conduction in the
bed/ walls and chemical reactions. The bed and freeboard models were thus treated as
separate domains and coupling between them is handled explicitly. However, the geometry of
kiln was assumed to be axisymmetric in this work and therefore the boundary conditions
were applied only in an approximate manner. Moreover, Mastorakos et al. (1999) assumed
coating formation throughout the kiln length. Karki et al. (2000) developed a 3D CFD based
model for simulating simultaneous combustion and heat transfer in cement kilns.
Karki et al. (2000) have used an effective thermal conductivity to define degree mixing in
bed region, developing a single computational model for simulating cement kilns. Different
values of effective thermal conductivities at different locations in the kiln were used.
However, there are no proper guidelines to choose proper effective thermal conductivity and
the values used are based on experience. Moreover, Karki et al. (2000) have focused on fluid
dynamics and heat transfer and did not model clinkerization reactions. The model predictions
thus gave only qualitative agreement with industrial observations reasonable for global
quantities of interest. Kolyfetis and Markatos (1996) also focused on coal combustion and
heat transfer in freeboard region and did not account clinker reactions. Also, Kolyfetis and
Markatos (1996) did not consider any coating formation in their computational model. The
above discussion clearly indicates that CFD based models presented earlier for cement kilns
are more or less approximate and not rigorous models. A comprehensive model for cement
kilns should account for simultaneous flow, heat transfer and reactions occurring in rotary
cement kilns.
It is also important to note that along with physical issues that needs to be captured, there are
numerical issues involved in cement kiln modeling. The free board region of the kiln in
156
which combustion of coal takes place and the bed region of the kiln where clinkerization
reactions take place are strongly coupled with each other. However, the characteristic time
and space scales of free board region and bed region are significantly different. The typical
velocities of gas phase in the free board region is about 10 m/s while typical velocities of
solid particles in the bed region is of the order of a typically about 0.05 m/s. In addition to the
time scales of the free board and bed region, actual chemical reactions occur at the scale of
the particles, which is much smaller than length scale of a kiln. Considering these vastly
different characteristic time and length scales of processes, building a single computational
fluid dynamics (CFD) based model with an objective of capturing all the relevant space and
time scales might fail it can become computationally extremely expensive. Earlier, Spang
(1972) has also reported convergence difficulties while solving flow, heat transfer and
reactions in the freeboard and the bed region simultaneously. Therefore in the present work,
we develop separate but coupled models for the bed and the freeboard regions. The strategy
was to simulate different regions with similar time scales separately and then couple them by
mass and energy communication via common boundaries. The methodology adopted will be
explained in detail in the next section. Before that we discuss the individual model equations
for bed and freeboard in the following.
6.2.2 Bed model

Several solid-solid and liquid-solid reactions occur in the bed region of rotary cement kilns.
The five major reactions considered by most of the researchers are given in Table 1. It can be
seen that these reactions involve components like CaCO3, CaO, C2S, C3S, C3A and C4AF.
While formation of some of the minor phases viz. C12A7, C2AS, CS, C3S2, CS2, CF, C2F etc.
have been reported (Hewlett, 1998), they are generally present in insignificant amount and
hence are usually neglected for modeling purposes (Spang 1972; Mastorakos et al.,1999). We
also consider only five major reactions occurring in cement kilns during clinker formation (as
given in Table 6.1) in the present study. One of the important issues in modeling these
reactions is the availability of the relevant kinetics. In Chapter 2 of the thesis, we have
critically analyzed the modeling of clinkerization reactions in cement kilns. The kinetic
parameters used in this work for clinker formation reactions were shown to work reasonably
well for 3 different industrial kilns covering wide range of operational conditions. Therefore,
these parameters were used to model kinetics in the present work
157
Table 6.1: Reactions, kinetics and heat of reactions (Refer Table 3.2, Chapter 3)
Reaction k0 E ∆H
(kJ/mol) (kJ/mol)
Bed
1. CaCO3 = CaO + CO2 1.18 × 10 3 (kmol/m2-s) 185 179.4
2. 2CaO + SiO2 = C2S 1.0 × 10 7 (m3/kg-s) 240 -127.6
3. C2S + CaO = C3S 1.0 × 10 9 (m3/kg-s) 420 16.0
4. 3CaO + Al2O3 = C3A 1.0 × 10 8 (m3/kg-s) 310 21.8
5. 4CaO + Al2O3 + Fe2O4 = C4AF 1.0 × 10 8 (m6/kg2-s) 330 -41.3
In our one-dimensional model of rotary cement kiln, the height variation of bed along the kiln
length was modeled using Kramer’s model (See for example, Chapter 3 of the thesis). The
same bed height profile can in principle be included in the CFD model. However, since the
focus was on understanding flame characteristics, the bed height variation along the kiln was
ignored in the present work to simplify the grid generation.
The motion of solids in rotating cylinder exhibits complex motion and various flow regimes.
Attempts have been made to simulate motion of solids in transverse section of rotating
cylinders as discussed in Chapter 2. The state of the art CFD models developed in Chapter 5
did not capture the details of solids motion accurately. These preliminary studies on
simulations of motion of solids in transverse plane of rotating cylinder however, indicated
that the flow generated in transverse section of kiln might be approximated by treating solids
as pseudo homogeneous fluids. The constant viscosity as well as power law viscosity model
was evaluated. Based on these studies, in this work we have treated solids as pseudo
homogeneous fluids with constant viscosity (2 kg/m-s).
Another important aspect of cement kilns is the formation of coating which has significant
influence on shell temperature profile and heat transfer. Formation of coating in cement kilns
is a complex function of liquid formation in bed, flow/reactions occurring in bed and
freeboard regions, kiln RPM and energy exchange between bed and freeboard. The
phenomenon of coating formation in cement kilns is not yet well understood. In the reaction
engineering based model developed in Chapter 3, the location of coating formation was
calculated as a part of the solution of model equations. Nevertheless, such an approach is
158
difficult to use with the CFD framework since it will involve dynamic meshing. The model
predictions for three industrial kilns with varying operating conditions and dimensions
presented in Chapter 3, showed that, the coating formation occurred only after half of the kiln
length from solids entry. This also seems to be consistent with shell temperature
measurements reported by Kolyfetis and Markatos (1996). However there is no data/model
which gives inputs regarding beginning of coating formation in cement kilns. In absence of
any information, the location of coating formation was assumed to be formed at 50 % of kiln
length from the feed end.
It was essential to formulate appropriate boundary conditions for the kiln walls to estimate
the heat losses through the kiln shell. The outer wall of the kiln was given a rotational
boundary condition (5.5 RPM, see Table 3) and heat loss through the shell was modeled
using convection and radiation. Ambient temperature was assumed to be 300 K. The
convective heat transfer coefficient was considered to be 30 W/m2 K as suggested by
Mastorakos et al. (1999) for industrial cement kilns. Mass, momentum and energy
conservation equations for the bed region are given below.

∂ρ
+ ∇ ⋅ ( ρu ) = S m , b (6.1)
∂t
Momentum conservation equation:
∂ ( ρu )
+ ∇ ⋅ ( ρuu ) = −∇p − ∇ ⋅ σ + ρg (6.2)
∂t
Energy conservation equation:
∂( ρE ) (6.3)
+ ∇ ⋅ ( u( ρE + p )) = ∇ ⋅ ( keff ∇T − Σh j J j + ( τ eff u )) + Se ,b
∂t
Species conservation equation:

∂ ( ρYi )
+ ∇ ⋅ ( ρuYi ) = ∇ ⋅ ( ρDi∇Yi ) + Ri (6.4)
∂t
In the above equations u is velocity, ρ is density, T is temperature and Yi is mass fraction of

species, i. Di is the diffusion coefficient of species i. The rate of reaction Ri was based on
simple Arrhenius law with activity coefficients and Energy of activation given in Table 6.1.
Sm, b and Se, b are the volumetric mass sink term (kg/m3-s) and volumetric heat sink term
159
(W/m3) respectively due to loss of CO2 due to calcination reaction occurring the bed. The
calculation of these terms is discussed later in this section. The equations were solved for
steady state so that time differentials in all the equations were zero.
6.2.3 Free board model

A computational model based on Eulerian-Lagrangian approach was developed to simulate a
pulverized coal burner in a cement kiln. Since the volume fraction of coal particles in the
freeboard region is not expected to go beyond 10 %, the motion and burning of coal particles
was modeled using the Lagrangian approach (Ranade, 2002). The standard k-ε model was
seen to predict the gas phase turbulence for freeboard region of cement kilns quite
satisfactorily and therefore was used in this work (Mastorakos et al., 1999; Karki et al, 2000).
Gas phase combustion was modeled using finite rate chemistry. Radiation was modeled using
the P-1 approach. The reason for choosing these models are discussed in the individual sub
sections later. The model equations for the continuous (Eulerian frame) and for the dispersed
particles (Lagrangian frame) are given below.
Continuous phase:
∂ρ
+ ∇ ⋅ ( ρv) = S m , comb + S m , calc (6.5)
∂t
where, Sm, comb (kg/m3s) is the mass added to the continuous phase from dispersed phase (i.e.
due to devolatilization, char reaction) due to coal combustion in freeboard region and Sm,calc
(kg/ m3s) is the mass added due to CO2 addition from bed due to calcination reaction.
Momentum Conservation:
∂ ( ρv )
+ ∇ ⋅ ( ρvv) = −∇p + ∇ ⋅ (τ ) + ρg + F (6.6)
∂t
where, p is the static pressure, τ is the stress tensor, ρg is the gravitational force and F is the
external body force that arises due to interaction of the dispersed phase and other model
dependent source terms.
Energy Equation:
∂( ρE )
+ ∇ ⋅ ( v( ρE + p )) = ∇ ⋅ ( keff ∇T − Σh j J j + ( τ eff v )) + S e ,calc + S e ,comb (6.7)
∂t
160
The first three terms on the right hand side of the equation represent heat transfer due to
conduction, species diffusion and viscous dissipation respectively. Se,comb represents the heat
source due to char combustion and Se,calc is the heat source due to CO2 addition from bed due
to calcination reaction.
Discrete phase:
The motion of the coal particles was simulated using the Lagrangian frame of reference by
solving the following equations
du p g x (ρp − ρ )
= FD (u − u p ) + + Fx (6.8)
dt ρp
where uP is a particle velocity and u is a continuous phase velocity. The additional force Fx
includes the forces on the particles that arise due to rotation of the reference frame. FD(u-up)
is the drag force per unit mass of the particle. The present computational model assumes coal
particles to be comprised of char, volatiles and ash. The computational models for coal
devolatilization, gas phase mixing and combustion used in Chapter 3 were shown to predict
the performance of three different kilns with different coal composition and calorific values
quite reasonably. Therefore, the same models were used in the present work. The coal
particle first gets heated by absorbing energy from the gas. The energy is received by
convection and radiation. Due to heat up the coal particle starts devolatilization. Some
approaches use correlations of volatile yield with particle temperature or define
devolatilization rates using single or two step Arrehenius schemes (Karki et al., 2000). The
present model assumes a constant rate devolatilization for coal combustion (Pillai, 1981). The
devolatilization constant recommended as 12.0 for coal combustion (Pillai, 1981) was used in
the simulations. The coal particle density was assumed to decrease in proportion to the
volatiles released in the gas phase (Heidenreich et al., 1999). On devolatilization, gas phase
combustion occurs. The investigations of coal devolatilization have given rise to a number of
models for volatiles and kinetics of their gas phase combustion (see for example, Li et al.,
2003). In many cases, assumption of a generalized single global step reaction of volatile fuel
(and assumption of CH4 as volatile matter in coal) have reproduced temperature profiles for
gas in freeboard region reasonably well for commercial coal combustors (Li et al., 2003; Guo
et al., 2003; Chen et al., 2001a; Chen et al., 2001b). In Chapter 3, it was shown that
simulation of three different industrial cement kilns fed with coal of different physical
161
compositions also lends an indirect justification for using the assumption of single step gas
phase combustion reaction. Hence we have used a global single step reaction of volatile fuel
to model gas phase combustion in this work as:
CH 4 + 2O 2 → CO2 + 2 H 2 O (6.9)
However, the framework developed is general enough that more complicated volatile
combustion reactions can be included as required. The kinetics and heat of reaction for the
assumed gas phase reaction are given in Table 6.1. It should be noted that the characteristic
reaction time scales at temperatures prevailing in kilns (>1500 K) is less than 1 ms.
Therefore, at such high temperatures, the reaction will be limited by turbulent mixing rather
than reaction kinetics. Two approaches i.e. generalized finite rate formulation [eddy breakup
(EBU) model] and mixture fraction/PDF formation are used generally to model mixing
limited turbulent gas phase combustion (Eaton et al., 1999). The main advantage of EBU
combustion model of Magnussen–Hjertager (1976) lies in its simplicity and robustness for
wide range of operating conditions. Hence it is widely used for predicting gas phase
combustion in coal combustors (Karki et al, 2000, Guo et al., 2003; Li et al., 2003 and
references cited there in) and therefore was used in this work. This model compares the
kinetic rate and the turbulence mixing rate and selects the lower rate for further calculations
as:
Rcombg = min (RS, EBU, RS, Arr) (6.10)
where
k y ox
RS , EBU = C R ρ min( y F , ),
ε b (6.11)
E
RS , Arr = Bs ρ y F y OX
2
exp(− s )
RT
yF is the mass fraction of the fuel and yOX is the mass fraction of oxygen. The values for
Arrhenius kinetic constants BS and ES are specified in Table 1. CR is a parameter of the eddy
breakup (EBU) turbulent combustion model, which is usually set to 4. Char combustion was
modeled using diffusion limited surface reaction model derived from model of Baum and
Street (1971). This model is extensively used for modeling char combustion in coal-fired
furnaces (Williams et al., 2002) and is therefore used in the present work. The model assumes
that the surface reaction proceeds at a rate determined by the diffusion of the gaseous oxidant
to the surface of the particle ignoring the kinetic contribution to the surface reaction rate.
162
Radiation modeling:
The radiation was modeled using the P-1 radiation model. The P1 radiation model is
recommended to use when the optical thickness aL >1, where ‘a’ is the absorption coefficient
of gas and L is the characteristic length of the system (Fluent 6.2 user manual). For the
operating conditions prevailing in the cement kilns, the optical thickness is generally > 1
(~ 1.2 to 1.8 for kiln under study) and therefore P1 model was used to model radiation in the
present work. P1 model is the simplest case of more general P-N model, which is based on
the expansion of the radiation intensity into an orthogonal series of spherical harmonics.
When only four terms in the series are used, the radiation heat flux is given by:
qr = −Γ∇G (6.12)
1
where, Γ = (6.13)
3(a + σ s ) − Cσ s
in which G is the incident radiation, a is the absorption coefficient, σs is the scattering

coefficient, C is the linear anisotropic phase function coefficient
From the transport equation of G and combining the above equation, we get,
− ∇qr = aG − 4aσT 4 (6.14)
When the effect of the coal particles in the P-1 radiation model is included, the incident
radiation is written as,
aσ T 4
− ∇q r = (a + a p )G − 4π ( + Ep ) (6.15)
π
where, ap is the equivalent absorption coefficient and Ep is the equivalent emission of the
particles. The local absorption coefficient for the gas-particle mixture was calculated using
the following expression (Boyd and Kent, 1986; Karki et al., 2000)
-Tg
a = 0.32 + 0.28 e 1135

(6.16)
where T is the temperature in Kelvin. In the vicinity of the burner region, the absorption
coefficient was augmented by 0.4 m-1 to account for soot (Boyd and Kent, 1986; Karki et al.,
2000). A constant value of 0.13 m-1 was used for isotropic scattering coefficient (Boyd and
Kent, 1986; Karki et al., 2000).
163
6.2.4 Mass transfer from bed to freeboard
As calcination reaction takes place in the bed region, CO2 is released from bed to the
freeboard region. The release of CO2 from bed to freeboard was modeled in each region
individually. For the bed region volumetric sink terms for mass (kg/m3 s) and energy (W/m3)
were applied to each cell through user defined functions (UDF’S) as
Sm,b = - RCO2 (6.17)
Where RCO2 is rate of CO2 produced in each cell and Sm,b (kg/m3 s) is the volumetric mass
sink in each cell (in Equation 1). The mass sink in each cell was thus calculated proportional
to amount of CO2 produced due to calcination reaction. The corresponding heat sink in each
cell was calculated as
Se,b = -RCO2 × CP, CO2 × (TCO2, avg – TBase) (6.18)
Where Se, b (W/m3) is the volumetric heat sink in each cell (see Equation 6.3). As a first
approximation we use an average temperature of bed in calcination region (i.e. TCO2, avg =
1000 K) to evaluate the heat sink from the bed. This mass and sink were passed as source to
the freeboard region as described below.
In this model, we implemented the transfer of CO2 from the bed to the freeboard as a function
of axial position. It is possible to transfer CO2 lost from the bed to the freeboard at every cell
interfaces appearing on the bed-freeboard boundary (cell to cell coupling or 2D coupling).
CO2 transfer from bed to freeboard can also implemented in the CFD model by averaging
over the chord of the bed at every axial position (1D coupling as done in Chapter 3). Since
the difference in the results predicted with the 1D and 2D coupling was not significant (as
will be discussed later) we used 1D coupling for mass transfer from bed to freeboard. From
the bed run the averaged loss of mass at different axial locations was calculated. From this
information mass flux Fz (kg/m2 s) at different axial locations could be calculated. The mass
source was passed to freeboard region as
L
S m ,calc = ∫ Fz ( z ) / V dA (6.19)
0
where Sm, calc (kg/ m3s) is the mass source of CO2 to the freeboard, Fz (z) is the mass flux
profile obtained along the kiln length, V is the volume of cell and dA is the cross sectional
area through which CO2 enters the freeboard region. Corresponding heat source term was
164
written for freeboard model to account for heat accompanied by CO2 from the bed was
written as
L
S e ,calc = ∫ ( Fz ( z ) / V ) × C P ,CO 2 × (TCO2, avg - 298.0) dA (6.20)
0
where Se, fb (W) is the heat source due to CO2 entering the freeboard region and TCO2, avg is the
average temperature of CO2 as discussed previously. With these model equations, we now
discuss the methodology adopted to simulate rotary cement kiln.
6.2.5 Methodology adopted
Though separate computational models were used for the bed and the free board region, the
computational grid was generated for the whole kiln. Different components and the generated
grid are shown in Figure 6.1.
(a) (b) (c)
(d) (e)
Figure 6.1: Computational grid

a: kiln burner,12 axial air ports, coal air inlet (blue), swirl air inlet (green);
b: bed region grid; c: cross sectional view of kiln (kiln internal diameter: 3.14m)
d: Burner grid; e: freeboard region grid
165
The computational grid for the bed, free board and shell region was extracted from this single
computational grid. The commercial grid generation code GAMBIT 2.0 was used to model
the geometry and to generate a computational grid for the entire kiln. All the simulations
were carried out using commercial FLUENT 6.2.9 solver. The overall methodology is shown
schematically in Figure 6.2 and explained below
Heat Loss
(Radiation + convection)
CO2 from Calcination

Reaction in bed Conduction heat
transfer (walls)
Temperature
Freeboard (bc* to freeboard)
Heat Flux
(bc* to bed) Bed
common surface
(bed/freeboard) *boundary condition
Figure 6.2: Coupling methodology
The bed model was first simulated independently. A pseudo-homogeneous approximation

was used to simulate the motion of solids in the bed region. In the first phase, a constant
viscosity was specified for the bed material as discussed earlier. Simulation of the bed region
was initiated by assuming a heat flux at the bed-free board common interface (considered as
wall for these simulations). The bed flow and the reactions occurring were solved
simultaneously to get the temperature profile for the common interface between the free
board and bed regions. The predicted bed temperature was then used to initiate simulations of
the free board region. Combustion of coal particles in the free board region was modelled
using the Eulerian-Lagrangian approach. The free board was then simulated using
temperature profile from the bed run to get the heat flux profile for the common interface.
This heat flux was then passed as boundary condition to simulate the bed region again. This
process was continued till the temperature and the heat flux through the interface did not
changed with further iterations. For the initial simulations 1d coupling was initiated.
Thereafter some simulations were also carried with 2d coupling between freeboard and bed.
This is discussed later in results and discussion section. Exchange of CO2 from bed to
166
freeboard due to calcination was accounted by proper source/sink terms. Suitable under-
relaxation factors were identified for faster convergence of coupled solution.
6.3 Results and Discussion
The computational model described in the previous section was used to understand the details
of burning zone and overall behavior of rotary cement kilns. A hexahedral grid of 18,100
(~ 12 nodes in radial direction and 150 nodes in axial direction) and 5,85,000 cells (~ 35
nodes in radial direction and 150 nodes in axial direction) was used to simulate bed and
freeboard regions of kiln respectively. Care was taken to resolve grid near the burner region
finer in order to capture swirl laden high velocity flows produced by the kiln burner. Grids of
similar magnitude are found to be adequate for cement kiln simulations [See
Mastarokos et al. (1999) and references cited there in]. The mass, momentum and energy
balance equations were discretized using first order discritization scheme. During individual
simulations of bed and freeboard the temperatures at different locations were monitored. The
individual runs for bed and freeboard were simulated till the temperatures at these locations
did not changed by an error of 0.1 %. Typically one bed run required CPU time of ~ 5 hrs and
freeboard required CPU time of ~ 20 hrs to convergence with this grid on a dual processor
machine of 2 Ghz processor speed and 2GB memory. Before presenting the simulation results
we initially present results for coupling of bed and freeboard region in the following.
6.3.1 Coupling of bed and freeboard region

The computational model was used to simulate typical industrial kiln whose dimensions and
operating conditions are specified in Table 6.3 (See Chapter 3). The properties of gas and
solids used for these simulations are specified in Table 6.2. The bed and freeboard regions of
kiln were simulated separately and communicated via mass and energy communication
through common interface. To accelerate the convergence, typically for first few iterations 1d
coupling was done. For this 1d coupling the temperature and heat fluxes were averaged
radially at each axial location and were passed to freeboard and bed runs respectively (See
section 6.2.3). An under relaxation of 0.5 was used in these simulations. The iterative
procedure was carried till the temperature and heat flux of common interface did not changed
further with in an error of ± 1%. Once 1d coupling was converged 2d coupling was initiated.
For these simulations coupling the heat flux and temperature values between common
interface of bed and freeboard were not averaged in radial direction. A cell to cell
167
correspondence of temperature and heat flux at common interface was carried. Since 1d runs
were already converged, an under relaxation of 1 was used in 2d coupling for faster
convergence. The iterations with 2d coupling were carried till the bed temperature in the kiln
did not changed further within an error of ± 0.1%. Typically 10 iterations of 1d coupling and
3 iterations for 2d coupling were required for complete convergence.
Table 6.2: Physical properties used in simulations

Sr. Variable Value
No.
1. Bed Density, kg/m3 1046+
2. Bed Heat capacity, kJ/ kg K 1.088$
3. Bed Thermal conductivity, W/m K 0.5*
4. Gas heat capacity (J/kg K) (0.106 TG + 1173) !
5. Gas Viscosity (kg/m/s) (0.1672 × 10 −5 TG − 1.058 × 10 −5 )!
*
Perry (1984); +Karki et al. (2000); $ Spang (1972);!Guo et al. (2003) ; ^
Table 6.3: Dimensions and operating condition of kiln

Sr. Variable Industrial Kiln
No.
1. Length, m 50
2. Inner refractory diameter, m 3.4
3. Coating thickness, m 0.13
4. Speed of rotation, RPM 5.5
5. Solids flow in, kg/s 38.88
6. Height at solid entry, m 0.46
7. Secondary air 13.469 kg/s; 1373K
8. Axial air 1.17926 kg/s; 313 K
9. Coal air 0.943 kg/s; 328 K
8. Swirl air 0.5266 kg/s; 313 K
10. Coal Fr, kg/s 1.25
11. Char 0.64
12. Volatiles 0.27
13. Ash Content 0.09
14. Calorific value, kcal/kg coal 6200
15 Coal Particle size, microns 100
168
However, it was observed that difference in predicted results of converged 1d coupled
simulations and 2d coupled simulations were insignificant. Typical results are shown in
Figure 6.3. Figure 6.3a and Figure 6.3b shows the comparison of temperature contours for
converged 1d and 2d coupled simulations (The numerical values are not shown in any of
contour plots to maintain secrecy of industrial data). The predictions for area weighted
average temperature of the gas in freeboard region are shown in Figure 6.3c. It can be seen
from Figure 6.3c the difference in averaged predictions is also insignificant. Moreover, the
percentage error in total amount of heat flux given to bed by the two methods is less than ~
0.5 %. Therefore, all simulations were carried using a 1d coupling methodology. This not
only saves CPU time of approximately ~ 20 % but also avoid convergence problems of
bed/freeboard coupling simulations (for example, a bed/freeboard coupling simulation with
direct 2d coupling has significant convergence problems).
Figure 6.3a: Temperature contours in central x plane (1D Coupled)
Figure 6.3b: Temperature contours in central x plane (2D Coupled)
169
2400
2200
1D Coupled
2D Coupled
2000
Gas Temperature, K
1800
1600
1400
1200
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized kiln length
Figure 6.3c: Comparison of area weighted averaged freeboard gas temperature
170
6.3.2 Simulations of rotary cement kilns
The converged simulations of bed and freeboard model were analyzed to gain insights into
flame characteristics and three-dimensional flow, temperature and composition distribution
with the kiln. Unfortunately there were no temperature and composition measurements
available inside the kiln. Hence the model predictions were initially compared with available
data at the kiln exit. The predicted results were also compared with those predicted by one
dimensional model presented in Chapter 3. A comparison of bed and freeboard temperature
predicted by the present CFD model (temperature predictions in central x-plane) and the 1d
model for the same kiln with same operating conditions are shown in Figure 6.4a and Figure
6.4b. A deviation in temperature predictions by CFD and 1d model, as can be seen from these
figures, could be due to difference in assumption of internal coating formation in both the
models. This is explained later in this section. The predicted mass fraction profiles of C3S and
C4AF along the kiln length by CFD and 1d model are shown in Figures 6.4c and 6.4d. These
results seem to be reasonable.
The temperature distribution of kiln wall controls the net loss to the environment which is
one of the major contributors in the net energy consumption in the kiln. Therefore it is
essential to predict the shell temperatures accurately. The comparison of shell temperatures
predicted by the two models is shown in Figure 6.5a. It is important to note that CFD model
was able to predict the temperature dip (~ 200 0 C) as observed in industrial data (see Figure
6.5b). This was due to additional resistance to heat transfer due to coating formation. Such
dips in shells temperatures have not been presented in previously published CFD models due
to inconsistent modeling of coating formation (Kolyfetis and Markatos, 1996; Mastorakos et
al., 1999). The quantitative difference in temperature predictions of the two models could be
due to difference in location of coating formation assumed/predicted by CFD and 1d model
respectively. 1d model was able to predict the coating formation based on operating
conditions (coating was predicted to be formed around 25 % of kiln length from the solid exit
end). However, in CFD approach, due to a prior meshing requirement, one has to assume
coating length in kiln before the simulations (coating formation was assumed be to formed
around 50 % of kiln length from solid exit in the present work).
The quantitative comparison of the temperature and clinker composition predicted by the
model with the industrial data and 1d model is given in Table 6.4. In general, the presented
171
CFD model was able to predict the overall behaviour of kiln reasonably when compared with
industrial observations and published results.
3000
3000
2500 CFD 2500
Temperature, K
1D Model
Temperature,
2000 2000
1500 1500
1000 1000
500 500
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized kiln length Normalized kiln length
Figure 6.4a: Comparison of bed temperature with Figure 6.4b: Comparison of freeboard temperature
results presented in Chapter 3 results presented in Chapter 3
1 1
CFD
C3S Mass Fraction
C4AF Mass fraction
0.8 0.8
1D Model CFD
0.6 0.6 1D
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized Kiln Length Normalized Kiln Length
Figure 6.4c: Comparison of clinker composition Figure 6.4d: Comparison of clinker composition
(C3S) results presented in Chapter 3 (C4AF) with results presented in Chapter 3
172
(6.5a) Predictions by computational model (6.5b) Reported shell temperature measurements from
Abscissa 1 corresponds to burner location industrial kiln (Kolyfetis and Markatos, 1996) Kiln Length:
75 m, Inner diameter: 4.35 m
Abscissa 0 corresponds to burner location
Figure 6.5: Comparison of shell temperatures with published literature
Table 6.4: Comparison of model predictions with industrial data

Sr. Variable Industrial CFD 1D (See, Table 3.8
No. data Chapter 3)
1. Mass Fraction C3S 0.483 0.506 0.481
2. Mass Fraction C2S 0.239 0.225 0.238
3. Mass Fraction C3A 0.051 0.052 0.052
4. Mass Fraction C4AF 0.143 0.129 0.146

5. Mass Fraction CaO 0.084 0.087 0.083
6. Temperature of solids, 1673 1632 1608
K
173
On obtaining reasonable agreement with industrial observations and previously published
results the CFD model was used to understand intricacies of burning zone of cement kilns
which cannot be predicted by 1D model. Typical predictions of the CFD model for coal
devolatilization, char burn out and CO2 mass fractions in the central plane of freeboard region
are shown in Figure 6.6. Information regarding location of release of volatiles in freeboard
region [See Figure 6.6a], char burn out [See Figure 6.6b] in the radial direction can be
obtained from such simulations. Such simulations can help to identify various zones of coal
combustion. The possibility of coal particles hitting the walls or bed due to improper burner
design or operating conditions can be figured out from such simulations. Figure 6.6c shows
CO2 mass fraction is central plane of freeboard regions. The CO2 mass fraction is higher near
the solid entrance due to calcinations reaction. Such information can be useful to identify
calcinations region in rotary kilns. Figure 6.6d represents the temperature contours in the
central plane of freeboard region. The Possibility of temperature zones beyond acceptable
limits due to tilting of flame and possible damages to the refractory can be examined using
such results. If the simulations indicate that the flame characteristics are not optimum, the
CFD model can be used to understand influence of various burner operating protocols on
flame characteristics. The figures predicted by developed model for temperature and mass
fraction predictions for coal and air in freeboard region seem to be reasonable when
compared with industrial observations. Thus the developed model was not only able to
predict the behavior of cement kilns as predicted by previously published results but also able
to provide information regarding burning zone intricacies quite satisfactorily.
174
Figure 6.6a: Coal devolatilisation, kg/s ( = 1.52 x 10-5; = 0)
Figure 6.6b: Char burn out, kg/s ( = 1.24 x 10-4; = 0)
Figure 6.6c: CO2 mass fraction ( = 1.0; = 0)
Figure 6.6d: Temperature contours in central vertical plane of freeboard

region, K
175
6.3.3 Effect of burner operational parameters on the kiln performance
On obtaining satisfactory predictions from the model, numerical experiments were carried
using the model to investigate influence of burner operational parameters on the kiln
behavior. The burner of cement kilns forms one of the most important components of the
entire system which controls the flame characteristics in the freeboard region and
subsequently the performance. Predetermined quantities of primary air and secondary air are
passed through the burners along with swirl air and fuel to get the desired flame
characteristics. For industrial cement kilns, typically primary air is passed through burner as
high momentum jets (~ 200-250 m/s) along with relatively slow secondary air (~ 15-20 m/s)
The swirl air passed through the burner opens the primary jet and influences the rate of
entrainment of secondary air in the primary air. The relationship between jet momentum and
the swirl/secondary component of air has significant effect on flame length and heat transfer
in cement kilns (Mullinger, 1984). It is observed that kilns with low jet momentum suffer
from poor air/fuel mixing and long flames leading to more fuel consumption (Mullinger,
1984). The developed computational model was used to analyse the effect of variation of
axial/swirl ratio through the burner (consequently effect of axial jet) on the flame
characteristics in freeboard region and corresponding effect on clinker composition.
The base case simulation (explained in Section 6.3.3) was carried out at an axial to swirl ratio
of 2 with a very high primary jet velocity (~ 245 m/s). To predict the freeboard region
temperature profiles at low momentum jets, the ratio of axial to swirl air was reduced from 2
to 1 (with axial velocity of reduced to ~ 100 m/s). The consequent change in the temperatures
due to reduced axial jet momentum in central plane of freeboard region is shown in Figure
6.7a and Figure 6.7b. As the primary air jet momentum was reduced the flame (shown by
temperature contours) tends to become larger (see Figure 6.8b). Consequently for low axial
jet simulation C3S composition decreased to 0.45 from 0.5 by reducing axial to swirl ratio
from 2 to 1. Both these results seem to be in reasonable agreement with previous results for
cement kiln coal burners (Mullinger, 1984). The reduction in axial velocity and consequent
increase in swirl component causes expansion of primary jet (i.e. enhanced jet entrainment,
see Figure 6.8a and Figure 6.8b) which is also in agreement with published results
(Frassoldati et al., 2005). Moreover, the temperature profiles obtained in freeboard region
(see Figure 6.8a and Figure 6.8b) for different axial/swirl ratios seems to be reasonable at
least qualitatively when compared with published results for burners operated with different
fuels (Huang and Yang, 2005). Thus, the developed computational model was able to capture
176
influence of ratio of swirl or axial air on the flame characteristics and the performance of
cement kilns.
Figure 6.7a: Temperature contours in central vertical plane of

freeboard region, K
Figure 6.7b: Temperature contours in central vertical plane of freeboard

region, K
The computational model was also analyzed to predict the temperature profiles in freeboard
region of the kiln for different burner operational parameters. For example, influence of
changing inlet coal particle size, coal flow rate, ash content of coal at same energy input (by
changing coal flow rate) and oxygen enrichment of the primary air on temperature profiles in
the kiln is shown in Figure 6.8a. It can be seen that, at the same levels of energy input by
coal, the temperatures are more or less similar (no significant influence of particle size or ash
content) except with oxygen-enriched air. As expected, decreasing the coal flow rate from 1.5
to 1 kg/s decreases the overall energy supplied to the kiln and therefore freeboard temperature
decreases. With coal particle containing more ash, the position of maximum temperature
shifts slightly away from the burner. When oxygen enriched air was used, significant increase
in the predicted maximum temperature was observed. This is because of enhanced
combustion rates due to availability of more oxygen due to oxygen enriched air. The
comparison of predicted mass fraction of volatiles for this case with the base case is shown in
Figure 6.8b. It can be seen that with enriched oxygen, the rate of consumption of volatiles is
significantly higher as expected. Thus the model can provide guidelines for ensuring the
177
desired flame size and position for different coal grades either by manipulating coal particle
size, swirl to axial air ratio or possible use of oxygen enrichment of primary air.
3500
3000
2500
Gas Temperature, K
2000
1500
Kiln-normal-operation
1000 Low-coal-flow-rate
Oxygen-enriched-air
Low-coal-Particle-size
500
Low-Coal-Grade
0
0 0.2 0.4 0.6 0.8 1
Kiln Length, m
Figure 6.8a: Predicted gas temperature profiles

(kiln center line; Abscissa 0 is burner end )
0.12
Volatiles Mass Fraction in Freeboard,
O2 enriched
0.1
Base
0.08
0.06
0.04
0.02
0
0 0.1 0.2 0.3 0.4
Kiln Length, m
Figure 6.8a: Predicted profiles for volatile mass fraction

(kiln center line; Abscissa 0 is burner end )
178
Further work on modelling of melting and using better constitutive equations for solids phase
is in progress and may enhance the utility of models and framework developed in this work.
6.4 Conclusions
A comprehensive framework of computational fluid dynamics based model was developed to
simulate coal combustion, radiation and clinkerization reactions occurring in rotary cement
kilns. Due to significantly different physical time scales, separate but coupled models were
developed for bed and freeboard regions of cement kilns. The Eulerian-Lagrangian based
model was developed for combustion of pulverized coal in the free board region of kilns. The
radiation was modelled using a P-1 radiation model. The solid-solid reactions in the bed
region were simulated using a pseudo-homogeneous model. The bed and the free board
region models were coupled together by mass and energy communication though common
boundaries. Some specific conclusions drawn from the present study are
• The presented coupling methodology to solve model equations representing processes
occurring in rotary cement kiln worked well and can be extended to other similar
complex reactors. It was observed that averaged coupling methodology (i.e. 1d
coupling) saves computational resources compared to the cell to cell coupling
methodology (i.e. 2d coupling) without jeopardizing accuracy of predicted results.
• The developed CFD model was able to predict the performance of cement kiln in
terms of the data at the exit of industrial kilns quite satisfactorily.
• The predicted results indicate that reducing burner primary jet velocities produce
longer flames in freeboard region and consequently increase energy requirements of
rotary cement kilns.
Overall the developed model and presented results will help in providing developing better
understanding of freeboard region and providing some clues for better burner operations.
179
7. Conclusions
In the present work, framework of realistic computational models was developed for
clinker manufacture in cement industry. The computational models were developed using
conventional reaction engineering approach and computational fluid dynamics based
approach. A software viz., Rotary Cement Kiln Simulator (RoCKS), was also developed
to capture influence of key design (kiln dimensions) and operating parameters (RPM,
solids flow rate, composition, coal flow rate and so on) on overall performance of cement
kilns. Key aspects of the developed models and scope of future work is discussed below.
1. Reaction engineering based model for rotary cement kilns

A model based on one-dimensional framework was developed to simulate rotary cement
kilns, considering most of the underlying physics for the first time. Models for simulating
reactions and heat transfer in the bed region and coal combustion and heat transfer in the
freeboard region were developed based on a segregated approach. Melting of solids and
coating formation within the kiln was accounted. Coal combustion in freeboard region
was modeled by considering devolatilisation, gas phase combustion and char reaction.
The model was used to simulate performance of three industrial kilns having wide range
of operating conditions. The model predictions agreed reasonably with the available plant
data for all the three kilns. The model was also able to predict influence of key design and
operating parameters like kiln tilt, kiln RPM on the overall kiln performance. The model
was also used to propose and to evaluate a practical solution of using a secondary shell to
reduce energy consumption in rotary cement kilns. Specific results obtained from model
simulations are briefed below:
• The simulations indicated that the operation of kiln with higher solid loading and
lower RPM is favourable from the point of view of energy consumption per kg of
clinker. Other operational concerns like increase in dusting and mixing however
need to be considered while identifying maximum solids flow rate or optimum
RPM for any specific rotary cement kiln.
• The use of a secondary shell with air stream passing through the gap between the
kiln shell and the secondary shell appears to be a promising method to reduce
180
energy consumption in rotary cement kilns. It is possible to recover ~ 80 % of the
energy lost to the surroundings (typically 150 kJ/kg clinker) by judicious choice
of secondary shell and air flow rate without degrading the quality of recovered
energy significantly.
2. RoCKS (Rotary cement kiln simulator): An Integrated reaction engineering

model for pre-heater, calciner, kiln and cooler in clinker manufacture
A comprehensive reaction engineering model was developed to simulate complex
processes occurring in clinker formation in cement industry. The models for pre-heater
and calciner were developed assuming solids and gas to be completely back mixed. The
computational model for the kiln was developed assuming gas and solids as plug flow.
The integrated simulator was converted into simple to use GUI based software for cement
industry, named as Rotary Cement Kiln Simulator (RoCKS). RoCKS was used to
simulate performance of pre-heater, calciner, kiln and cooler for clinker formation. These
results were satisfactory when compared with available plant data and industrial
observations. Specific conclusions based on this model are:
• Including an additional pre-heater reduces net energy consumption. The
developed model can be used to evaluate relative benefits of energy savings by
additional pre-heater and required additional capital expenses.
• There is an optimum value for percentage of calcination carried out in calciner
with respect to overall energy consumption in clinker manufacture. With the
parameters selected in this work, this optimum value of percentage calcination in
calciner is about 70.
• The simulation results indicated that operating kiln with higher solid loading,
lower RPM, lower tilt and lower grate speed reduces energy consumption per
unit production. The upper limit on solid loading (bed height) and lower limits on
RPM and tilt (mixing and heat transfer) need to be identified based on other
practical issues.
181
3. CFD model for rotary cement kilns
A comprehensive framework of computational fluid dynamics based model was
developed to simulate transport processes in rotary cement kilns. Initially CFD models
were developed for predicting solids motion in transverse section of rotary kilns. These
models illustrated that use of simpler models like pseudo-homogeneous assumption for
the solids phase were found to give results similar to those as predicted by more
complicated models. The solids in the bed region of cement kilns were therefore
simulated using a pseudo-homogeneous assumption. The Eulerian-Lagrangian based
model was developed for combustion of pulverized coal in the free board region of kilns.
The radiation was modelled using a P-1 radiation model. The bed and the free board
region models were solved separately and coupled together via mass and energy transfer
though common boundaries. Specific conclusions drawn from this study are:
• The developed coupling methodology is an efficient way for modeling rotary
cement kilns which involves physical processes having significantly different
time scales. Coupling of bed and freeboard regions for cement kilns using
averaged coupling (i.e. 1d coupling) seems to be a better choice over cell to cell
(i.e. 2d coupling) since the former not only saves computational time by ~ 20 %
but also avoids convergence difficulties.
• The developed CFD model was able to predict the performance of cement kiln in
terms of the data at the exit of industrial kilns and published literature quite
satisfactorily.
• The predicted results indicate that reducing burner primary jet velocities produce
longer flames in freeboard region and consequently increase energy requirements
of rotary cement kilns.
On the whole, the developed models and presented results will be helpful for developing
better understanding and for realizing performance enhancements for clinker
manufacturing in cement industry.
182
4. Scope for future work
• Developing accurate frictional viscosity models for flow of solids in transverse
plane of the kiln is essential to make this framework more comprehensive
• The effect of particle size distribution is neglected in the scope of this work.
Incorporating size distribution of raw meal particles would be the next milestone
in further development of this framework.
• Incorporating population balance models to capture size of particles due to
agglomeration after melting of solids is another aspect which can be added in the
present framework of cement kiln models.
• The internal conduction inside clinker particles is not considered while
developing computational model for grate coolers. The present framework can be
extended by developing a more comprehensive and computationally intense
model for clinker cooler considering the internal gradients in the particles in
clinker cooler.
183
References
Barr, P. V., Brimacombe, J. K. and Watkinson, A. P., 1989, “A heat transfer model for the
rotary kiln- I. Pilot kiln trials”, Met. Trans B., 20B. 391-402.
Boateng, A. A., 1998, “Boundary layer modelling of granular flow in the transverse plane of
a partially filled rotating cylinder”, International Journal of Multiphase Flow, 24, 499-521.
Boateng, A. A. and Barr, P. V., 1997, “Granular flow behaviour in the transverse plane of a
partially filled rotating cylinder”, Journal of Fluid Mechanics, 330, 233-249.
Boateng, A. A. and Barr, P. V., 1996, “A thermal model for the rotary kiln including heat
transfer within the bed”, International Journal of Heat and Mass Transfer, 39, 2131-2147.
Borgwardt, R. H., 1985, “Calcination kinetics and surface area of dispersed limestone
particles”, AIChE Journal, 31, 103
Beretka, J. and Brown, T., 1983, “Effect of particle size on the kinetics of the reaction
between Magnesium and Aluminium Oxide”, J. Amer. Cerm. Soc., 66, 383-388.
Boyd, R. K. and Kent, J. H., 1986, “Three-dimensional furnace computer modeling”. 21st
Symp. (Int.) Combust., [Proc.], 265.
Chen, C., Horio, M. and Kojima, T., 2001a, “Use of numerical modeling in the design and
scale up of entrained flow coal gasifiers”, Fuel, 80, 1513-1523
Chen, K.S., Hsu, W.T., Lin, Y.C., Ho, Y.T. and Wu, C.H., 2001b, “Combustion Modeling
and Performance Evaluation in a full-scale rotary kiln incinerator”, J. Air & Waste Manage.
Assoc., 51, 885-894
Chern, J. and Hayhurst, A.N., 2004, “Does a large coal particle in a hot fluidized bed lose its
volatile content according to shrinking core model ?” Comb and Flame, 139, 208-221.
184
Cleary, P. W., Metcalfe, G. and Liffman, K., 1998, “How well do discrete element granular
flow models capture the essentials of mixing processes?” Applied Mathematical Modelling,
22, 995-1008.
Dhanjal, S.K., Barr, P.V. and Watkinson, A.P., 2004, “The rotary kiln: An investigation of
bed heat transfer in transverse plane”, Met. And Mat. Trans B, 35B, 1059-.
Ding, Y. L., Seville, J. P. K., Forster, R. and Parker, D. J., 2001, “Solids motion in rolling
mode rotating drums operated at low to medium rotational speeds”, Chemical Engineering
Science, 56, 1769-1780
Duchesne, C., Thibault, J. and Bazin, C., 1996, “Modelling of the solids transportation within
an industrial rotary dryer: A simple model”, Industrial Engineering Chemistry Research, 35,
2334-2341.
Djuric, M. and Ranogajec J., 2002, “Mathematical modeling in cement technology”, in -

Advances in Cement Technology: Chemistry, Manufacture and Testing, 2nd ed. Editor - S.N.
Ghosh, New Delhi, Tech Books International, India
Eaton, A.M., Smoot, L.D., Hill, S.C., and Eatough, C.N., 1999, “Components, formulations,
solutions, evaluation, and application of comprehensive combustion models” Prog. Energy
Comb. Sci., 25, 387-436.
Engin, T. and Ari, V., 2005, “Energy auditing and recovery for dry type cement rotary kiln
systems–A case study”, Energy Conversion and Management, Volume 46, Issue 4, 551-562
Friedman, S.J. and Marshall, W.R., 1949, “Studies in rotary drying, Part- I, Hold up and
dusting”, Chem. Engg. Prog., 45(8), 482.
Ferron, J. and Singh, D., 1991, “Rotary kiln transport processes”, A.I.Ch.E. Journal, 37, 747-
758.
185
Frassoldati, A, Frigerioa, S., Colombob, E., Inzolib, F. and Faravelli, T, 2005,
“Determination of NOx emissions from strong swirling confined flames with an integrated
CFD-based procedure”, Chemical Engineering Science, 60, 2851-2869.
Garcya-Labiano, F., Abad, L.F. de Diego, Gayan, P. and Adanez, J., 2002, “Calcination of
calcium-based sorbents at pressure in a broad range of CO2 concentrations”, Chem. Eng.
Sci., 57, 2381– 2393.
Gang, L, Rongxian, L. and Zhou, L., 2000, “Numerical simulation of gas particle flows with
different swirl numbers in a swirl burner” Tsinghua Science And Technology, Vol 5, No 1,
96-99.
Gorog, J. P., Brimacombe, J. K. and Adams, T. N., 1981, “Radiative heat transfer in rotary
kilns”, Met. Trans B, 12B, 55-69.
Gorog, J. P., Brimacombe, J. K. and Adams, T. N., 1983, “Heat transfer from flames in a
rotary kiln”, Met. Trans B, 14B, 411-424.
Guo, Y.C., Chan, C.K. and Lau, K.S., 2003, “Numerical studies of pulverized coal
combustion in a tubular coal combustor with slanted oxygen jet”, Fuel, 82, 893–907
Gupta, A.V.S.S.K.S. and Nag, P.K., 2000, “Prediction of heat transfer coefficient in the
cyclone separator of a CFB”, Int. J. Energy Res. 24, 1064-1079.
Hamor, R.J., Smith, I.W. and Tyler, R.J., 1973, “Kinetics of combustion of pulverised brown
coal char between 630 and 2200 K”, Combust. Flame 21, 153.
Hao, Y. J. and Tanaka, T., 1990, “Analysis of solid-solid reaction controlled by

unideirectional diffusion”, Int. Chem. Eng. (English Trans.), 30, 244-253.
Hewlett, P.C., 1998, Lea’s Chemistry of Cement and Concrete, Arnold, London.
186
Heidenreich, C.A., Yan, H.M. and Zhang D.K., 1999, “Mathematical modelling of pyrolysis
of large coal particles—estimation of kinetic parameters for methane evolution”, Fuel, 78,
557–566
Heinen, H., Brimacombe, J. K. and Watkinson, A. P., 1983, “Experimental studies of

transverse bed motion in rotary kilns”, Meteorological Transactions B, 14B, 191-205.
Hills, A.W.D., 1968, “The mechanism of the thermal decomposition of calcium carbonate”,
Chem. Eng. Sci., 23, 297-320
Hindmarsh, A.C., 1983, ODEPACK, a systematized collection of ODE solvers. In:

Stepleman, R.S., et al. (Eds.), Scienti.c Computing. North-Holland, Amsterdam, 1983.
IMACS Transactions on Scientific Computation 1, 55–64.
Hottel, H.C., 1967, Radiative heat transfer, 1st Edition, New York: McGraw-Hill.
Huang, Y. and Yang, V., 2005, “Effect of swirl on combustion dynamics in a lean-premixed
swirl-stabilized combustor”, Proceedings of the Combustion Institute, 30, 1775–1782
Ingraham, T. and Marier P., 1963, “Kinetic studies on the thermal decomposition of calcium
carbonate,” Canadian Journal of Chemical Engineering, 63, 170–173.
Irfan, A. and Dogu, G., 2001, “Calcination kinetics of high purity limestones”, Chemical
Engineering Journal, 83, 131-137.
Johansen, V., 1973, “Model for the reaction between particles and portland cement clinker,”
J. Amer. Cerm. Soc., 56(9), 450–454.
Kang, S.K., Kwon, T.W. and Kim, S.D., 1989, “Hydrodynamic characteristics of cyclone
Reactors”, Powder Technology 58,211-220.
Karki, K.C., Patankar, S.V. and Grant, J., 2000, “Simulation of fluid flow, combustion and
heat transfer in a coal-fired cement kiln”, FACT-Vol 23/ HTD-Vol. 367, Combustion, Fire
and Computational Modeling of Industrial Combustion Systems, ASME.
187
Khakhar, D. V., McCarthy, J. J., Sninbrot, T. and Ottino J. M., 1997, “Transverse flow and
mixing of granular material in a rotating cylinder”, Phys. Fluids, 9(1), 31-43.
Khinast, J., Krammer, G.F., Brunner, C. and Staudinger, G., 1996, “Decomposition of
limestone: the influence of CO2 and particle size on the reaction rate”, Chem. Eng. Sci., 51,
623–634.
Khraisha, Y.H. and Dugwell, D. R., 1992, “Coal combustion and limestone calcination in a
suspension reactor”, Chem. Eng. Sci., 47, 993–1006.
Kolyfetis, E. and Markatos. N. C.,1996, “Aerodynamics and coal – Flame modeling in

burning zone of cement rotary kilns, part 2. , ZKG International, 49, 326-334
Kramers, H. and Croockewit, P., 1952, “The passage of granular solids through inclined
rotary kilns”, Chemical Engineering Science, 1, 259–265.
Lebas, E., Hanrot, F., Ablitzer, D. and Houzelot, J.L., 1995, “Experimental study of residence
time, particle movement and bed depth profile in rotary kilns”, Can J. Chem. Engg., 73, 173-
180.
Lede, J., Li, H.Z., Soulalignac, F. and Villermaux, J., 1987, “Measurment of solid particle
residence time in cyclone reactor: A comparison of four methods”, Chem. Eng. Process, 22,
215-222.
Lee, J., Keener, T.C., Knoderera, M. and Khang, S., 1993, “Thermal decomposition of
limestone in a large scale thermogravimetric analyzer”, Thermochemica Acta, 213, 223-240.
Li, Z.Q., Wei, F. and Y. Jin, 2003, “Numerical simulation of pulverized coal combustion and
NO formation”, Chemical Engineering Science, 58, 5161 – 5171
Li, S.Q., Yan, J.H., Li, R.D., Chi, Y. and Cen K.F., 2002, “Axial transport and residence time
of MSW in rotary kilns, Part – I. Experimental”, Powder Technology, 126, 217-227
Li, K.W., 1974, “Entrainment in rotary cylinders”, AIChE J., 20(5), 1031-1034.
188
Liu, X. Y. and Specht, E., 2006, “Mean residence time and hold up of solids in rotary kilns”,
Chemical Engineering Science, 61, 5176-5181
Locher, G., 2002, “Mathematical models for the cement clinker burning process, Part 3:
Rotary Kiln”, ZKG International, 55 (3), 68-80.
Lu, J., Huang, L., Hu, Z. and Wang, S., 2004, “Simulation of gas-solid, two-phase flow, coal
combustion and raw meal calcination in a pre-calciner”, ZKG International, 57(2) 55-63.
Martin, G., 1932, Cement engineering and thermodynamics applied to the cement rotary kiln.
The Technical Press, London.
Martins, M.A., Oliveira, L.S. and Franca, A.S., 2001, “Modeling and simulation of petroleum
coke calcination in rotary kilns”, Fuel, 80, 1611-1622.
Martins, M, A., Oliveira, L, S. and Franca, A.S., 2002, “Modeling and simulation of
Limestone calcinations in rotary kilns”, ZKG International. 55(4), 76-87.
Marias, F., 2003, “A model of a rotary kiln incinerator including processes occurring within
the solid and gaseous phase”, Computer and Chemical Engineering, 27, 813-825.
Mastorakos, E., Massias, A., Tsakiroglou, C.D., Goussis D.A. and Burganos V.N., 1999,
“CFD predictions for cement kiln including flame modeling, heat transfer and clinker
chemistry”, Applied Mathematical Modelling, 23, 55-76.
Magnussen, B.F and Hjertager, B. H., 1976, “On mathematical modeling of turbulent
combustion with special emphasis on soot formation and combustion”, 16th
symposium(International) on Combustion, 719-729.
Megalos, N.P., Smith, N.L. and Zhang, Z.L., 2001, “The potential for low NOx from a
precessing jet burner of coal”, Combustion and Flame 124:50–64
Mullinger, P.J., 1984, “Burner design for coal fired cement kilns”, World Cement,348-353.
189
Nsofor, E. C. and Adebiyi, G. A., 2001, “Measurements of the gas-particle convective heat
transfer coefficient in a packed bed for high-temperature energy storage”, Experimental
Thermal and Fluid Sciences, Vol. 24, No. 1-2, 1-9.
Parker, D. J., Dijkstra, A. E., Martin, T. W. and Seville, J. P. K., 1997, “Positron emission
particle tracking studies of spherical particle motion in rotating drums”, Chemical
Engineering Science, 52, 2011-2022.
Perron, J. and Bui, R. T., 1990, “Rotary cylinders: solid transport prediction by dimensional
and rheological analysis”, Canadian Journal of Chemical Engineering, 68, 61-68.
Peray, K. E., 1986, Rotary Cement kilns, 2nd Edition, Chemical Publishing Co. Ltd., NY.
Perry, R.H., 1984, Perry’s Chemical Engineers’ Handbook, 6th Edition, Don Green, Mc-Graw
Hill, New York.
Pillai, K. K., 1981, “The Influence of Coal Type on Devolatilization and Combustion in
Fluidized Beds.” J. Inst. Energy, 142.
Quirrenbach, F.J., 1960, Allg. Warmetch, (9), 271-276
Richman, M. W. and Oyediran, A. A. 1992, “Grain size reduction in granular fows of

spheres: the e!ects of critical impact energy”, .Journal of Applied Mechanics, 59, 17-22.
Ranade, V.V., 2002, Computational Flow Modeling for Chemical Reactor Engineering,
Academic Press, London.
Rao, T.R., Gunn, D.J. and Bowen, J.H., 1989, “Kinetics of calcium carbonate
decomposition”, Chem. Eng. Res. Des., 67, 38-47.
Ranz, W.E. and Marshall, Jr, 1952, “Evaporation from Drops, Part I. Chem. Eng. Prog.,
48(3):141-146.
190
Saeman, W. C., 1951, “Passage of solids through rotary kilns”, Chemical Engineering
Progress, 47, 508-514.
Satterfield, C.H. and Feakes, F., 1959, “Kinetics of the thermal decomposition of calcium
carbonate”, AIChE J., 5, 115-122.
Schaeffer, D. J., 1987, “Instability in the evolution equations describing incompressible

granular flow”, J. Differ. Equa., 66, 19– 50.
Selcuk, N., Batu, A. and Ayranci, I., 2002, “Performance of method of lines solution of
discrete ordinates method in the freeboard of a bubbling fluidized bed combustor”, Journal of
Quantitative Spectroscopy and Radiative Transfer, 73, 503–516
Sherritt, R. G., Chaouki, J., Mehrotra A.K. and Behie, L. A., 2003, “Axial dispersion in the
three-dimensional mixing of particles in a rotating drum reactor”, Chemical Engineering
Science, 58(2), 401-415.
Shrivastava, A. and Sundaresan, S., 2002, “Role of wall friction in fluidization and standpipe
flow”, Powder Technology, 124, 45-54.
Shrivastava, A. and Sundaresan, S., 2003, “Analysis of a frictional-kinetic model for gas
particle flow”, Powder Technology, 129, 72-85.
Smith, I.W., 1971, “Kinetics of combustion of size graded pulverized fuels in temperature
range of 1200 -2270 K”, Combustion and Flame, 17(3), 303-314
Spang, H. A., 1972, “A dynamic model of a cement kiln”, Automatica, 8, 309-323.
Spurling, R.J., Davidson, J.F. and Scott D.M., 2001, “The transient response of granular
flows in an inclined rotating cylinder”, Trans IchemE, 79, 51-61.
Solomon, P.R., Hamblen D.G., Carangelo, R.M., Serio M.A. and Deshpande G.V., “Very
rapid coal pyrolysis”, Fuel, 65, 182-194.
191
Stanmore, B.R. and Gilot, P., 2005, “Review—calcination and carbonation of limestone
during thermal cycling for CO2 sequestration”, Fuel Processing Technology, 86,1707– 1743
Syamlal, M, Rogers, W. and O' Brien T. J., (1993). MFIX Documentation: Volume 1, Theory
Guide. National Technical Information Service, Springfield, VA..DOE/METC-9411004,
NTIS/DE9400087.
Thunman, H. and Leckner, B., 2002, “ Thermal conductivity of wood – models for different
stages of combustion”, Biomass and Bioenergy, 23, 47-54.
Tscheng, S. H. and Watkinson, A. P., 1979, “Convective heat transfer on rotary kilns”, Can.
J. Chem. Engg., 57, 433-443.
Van der Lans, R.P., Glarborg, P. and Dam-Johansen, K., 1999, “Influence of process
parameters on nitrogen oxide formation in pulverized coal burners”, Prog. Energy Comb.
Sci., 23, 349-377.
Walton, O. R. and Braun, R. L., 1993, “Simulation of rotary-drum and repose tests for
frictional spheres and rigid sphere clusters”, Proceedings of Joint DOE/NSP Workshop on
Flow of Particulates and Fluids.
Watkinson, A. P. and Brimacombe, J. K., 1982, “Limestone calcination in a rotary kiln”, Met.
Trans. B, 13B, 369-378
Williams, A., Backreedy, R., Habib, R., Jones, J.M. and Pourkashanian, M., 2002, “Modeling
coal combustion: the current position”, Fuel, 81, 605-618
Vahl, L. and Kingma, W.G., 1952, “Transport of solids through horizontal rotary cylinders”,
Chem. Eng. Sci., 1, 253.
Young, B.C. and Smith, I.W., 1989, “The combustion of Loy Yang Brown coal char”,
Combustion and Flame, 76, 29-35.
192
Journal Publications
1. Mujumdar, K.S., Arora, A., and Ranade, V.V., Modeling of Rotary Cement Kilns: Applications to
Reduction in Energy Consumption. Ind. Eng. Chem. Res., 45, (7), 2315 -2330, 2006.
2. Kaustubh S. Mujumdar and Vivek V. Ranade, Simulation of Rotary Cement Kilns Using a One-
dimensional Model”, Chemical Engineering Research and Design, 84, 165-177, 2006.
3. Kaustubh S. Mujumdar, Ganesh K.V., Sarita B. Kulkarni and Vivek V. Ranade, “Rotary Cement
Kiln Simulator (RoCKS): Integrated Modeling of Pre-heater, Calciner, Kiln and Clinker Cooler”,
Chemical Engineering Science, 62, 2007, 2590-2607.
4. Mujumdar, K.S. and Ranade V.V. A CFD tour of a cement kiln. Chemical Engineer (London),
777, 39-41, 2006
5. Kaustubh S. Mujumdar and Vivek V. Ranade, “CFD Modeling of rotary cement kilns” accepted
for publication in Asia Pacific Journal of Chemical Engineering
Conference/Poster presentations
6. Sumedh Warudkar, V.Ganesh Kumar, Kaustubh S. Mujumdar and Vivek V. Ranade, A
Phenomenolgy based model for cement calciners, presented at CHEMCON-Delhi, 2005
5. Kaustubh S. Mujumdar and Vivek V. Ranade, CFD modeling of rotary cement kilns: Influence of
Characteristics of Coal Combustion on Clinker Composition, presented at CFD FOR CRE IV-
Italy, 2004.
6. Amit Arora, Kaustubh Mujumdar and Vivek V. Ranade, A one-dimensional model for Rotary
cement kilns, Presented at Chemcon-Mumbai, 2004 (Received Best Paper Presentation award)
7. Kaustubh S. Mujumdar and Vivek V. Ranade, Coupling of CFD models with different time scales
- A case study of rotary cement kiln, FLUENT UGM for India and South East Asia, Pune, 2003
(Received Best Technical Paper Award)
8. Vivek V. Ranade and Kaustubh S. Mujumdar, CFD simulations of solid motion in the transverse
plane of rotating kilns, Third International Conference on CFD in the Minerals and Process
Industries, CSIRO, Melbourne, Australia, 10-12 December 2003.
9. Kaustubh S. Mujumdar and Vivek V. Ranade, CFD simulations of heat transfer in transverse plane
of rotating kilns, CHEMCON- Bhubhaneshwar, 2003.
10. Vivek V. Ranade, Kaustubh S. Mujumdar, S. Bandopadhyay and Devaki Ravi Kumar, Modeling
of Rotary Cement Kiln, ISPSEC, I.I.T, Bombay, 2003
11. Kaustubh S. Mujumdar and Vivek V. Ranade, Modeling of Rotary Cement Kilns, CHEMCON-
Hyderabad, 2002
Acknowledgments
There are many people whom I owe acknowledgement and who have helped me in
several ways during my doctorate degree. I would like to thank my advisor Prof. Anurag
Mehra for his constant support throughout my work. Interactions with him during the
thesis work and particularly during his course “Reaction Engineering in Dispersed Phase
System” was combination of real fun plus deep knowledge building about Chemical
Engineering. I am also thankful to my RPC members Prof. A.K. Suresh and Prof. D.V.
Khakhar for their valuable suggestions during the RPC meetings.
I am highly indebted to Dr. V.V. Ranade, my supervisor at National Chemical Laboratory

(NCL), Pune. Working with him was one of the most fortunate things that have happened
in my life. I could not have imagined having a better advisor and mentor for my PhD, and
without his knowledge, perceptiveness and approach to crack problems I would never
have finished in time. With very high standards of quality work set by him in our group at
NCL, Pune I would like to repeat my colleague Dr. Vivek Buwa’s words “I hope this
work at least partly satisfies his expectations.”
There are several friends whom I would like to acknowledge whom I met during my stay
at our group Industrial Flow Modeling Group (IFMG) at NCL, Pune. I would like to
acknowledge my seniors Vivek, Gunjal and Khopkar who helped me in several things
other than technical stuff. Special thanks to one of my colleague Ranjeet for helping me
several times during my work. Interacting with Ranjeet has always helped me in getting
some new idea to tackle a problem. Also I would like to thank him for sitting with me
and helping me out in debugging some of my FORTAN codes several times. I would also
like to thank my other colleagues Mohan and Ajay for their constant help in some or the
other way. I would also like to thank several other friends Naren, Sarita, Dev, Madhavi,
Latif, Ganapathy, Prakash, Abhijit, Amit, Yogesh, Madhav, Subbu, Rashmi, Puneet,
Surendra, Himanshu, Manan, Gaurav Gupta for making my stay at IFMG very pleasant. I
would also like to thank Sandip Talwalkar from IIT, Bombay who has been a constant
source of help during my visits at IITB.
I would specially like to thank three of my colleagues Amit Arora, Sumedh and Ganesh
who worked with me during my stay at IFMG. I would like acknowledge all the three for
their enthusiasm, promptness towards work and the awareness of time frame for
completing the undertaken work in time.
I would like to acknowledge CSIR, India for providing the financial support during my
research work.
I would also like to say thanks to my entire family particularly to my mom and my dad
and most importantly to my wife Namrata for standing by me at all times, may be good or
bad. I would also like to thank God for blessing me and my wife with our baby kid
“Ananya” who has added some of very sweet memories that have happened during my
Ph.D.
Kaustubh S. Mujumdar

TH1638

Uploaded by

Copyright:

Available Formats

TH1638

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TH1638

Uploaded by

Copyright:

Available Formats

MODELING OF ROTARY CEMENT KILNS

Submitted in partial fulfillment of the requirements

Roll no. 03402701

DEPARTMENT OF CHEMICAL ENGINEERING

is approved for the degree of DOCTOR OF PHILOSOPHY

CERTICATE OF COURSE WORK

Sr. No Course No. Course Name Credits

I.I.T Bombay Dy. Registrar (Academic)

Acl cross sectional area covered by the charge, m2

Q '' Heat lost by gas to bed and walls, kJ/m s

Calcineous Exhaust to atmosphere

Pre-heater Hot gases

Calciner Kiln Exhaust

Figure 1.1: Schematic of clinker manufacture

1.2 Rotary Cement Kilns

Figure 1.2: Schematic of rotary cement kiln

1.3 Motivation of the Present Work and Methodology Adopted

Figure 1.3: Integrated modeling approach for rotary cement kilns

Reaction Engineering Model

Figure 1.4: Multi-layer modeling methodology for cement kilns

1.4 Organization of Thesis

Detailed literature review

Modeling of Rotary Cement Kilns

Reaction engineering Computational fluid dynamics

Chapter 3 and Chapter 4 Chapter 5 and Chapter 6

Figure 1.5: Organization of the thesis

Raw Meal Clinker

Decomposition and Solid Solid Solid Liquid

Figure 2.1: Key issues in modeling rotary cement kilns

Flow Heat Transfer

Figure 2.2: Schematic of various processes occurring in cement kilns

2.2 Flow of Solids in Rotary Cement Kilns

2.2.1 Flow of solids in axial direction

Attempts of experiments and computational models for understanding motion of solids in

Normalized Kiln Length (-) (m)

Normalised Kiln Length (-)

Figure 2.6 Effect of axial dispersion coefficient on the kiln performance

2.2.2 Flow of solids in transverse direction

Figure 2.7 Flow regimes in transverse section

Figure 2.8: Flow of solids in rolling regime

2.3 Clinkerization Reactions

Reaction 1 (Equation 2.3, calcination reaction) is a decomposition reaction. Reactions 2, 3

Reviewing literature of clinkerization kinetics survey suggests that calcination reaction

2.4 Melt and Coating Formation

2.5 Heat Transfer

Figure 2.10: Heat transfer in rotary cement kilns

2.6 Freeboard Region of Cement Kilns

Accurate modeling of freeboard region is a key in developing computational models for

2.7 Summary of Key Issues

• It is essential to simulate motion of solids/bed height in the bed region. Axial

2.8 Performance Models for Rotary Cement Kilns

2.8.1. One dimensional models

2.8.2 CFD Models

Publications based on this work

In this chapter it is attempted to develop comprehensive mathematical model to simulate

3.2 Computational Model and Solution Methodology

3.2.1 Model for calculating bed height in rotary kilns

Figure 3.0: Various parameters influencing bed height in rotating kilns

3.2.2 Bed model

Equation describing rate of reduction of radius of a single particle was re-written as an