International Journal of Heat and Mass Transfer 106 (2017) 263–279

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Numerical simulation of heat transfer during production of rutile

titanium dioxide in a rotary kiln
Ashish Agrawal, P.S. Ghoshdastidar ⇑
Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur, U.P. 208016, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a computational heat transfer model of a rotary kiln used for the production of rutile
Received 24 February 2016 titanium dioxide by the calcination of paste-like hydrous titanium dioxide. The work details the mod-
Received in revised form 4 October 2016 elling of several chemical reactions occurring in the solid bed region along with turbulent convection
Accepted 7 October 2016
of gas, radiation heat exchange among hot gas, refractory wall and the solid surface, and conduction in
the refractory wall. Finite-difference techniques are used and the steady state thermal conditions are
assumed. The kiln is divided into axial segments of equal length. The solution is of marching type and
Keywords:
proceeds from the solid inlet to the solid outlet. The direction of gas flow is opposite to that of the solids.
Heat transfer
Rotary kiln
Mass balance of each species in the solid charge, and mass and energy balances of the solid and gas in an
Rutile titanium dioxide axial segment are used to obtain solids and gas temperatures, and species concentration at the exit of that
Simulation segment. The kiln length predicted by the present model is 45.75 m as compared to 45 m of an actual kiln
reported by Ginsberg and Modigell (2011). The steady-state axial gas and solid temperature profiles have
been also satisfactorily validated with the numerical results of the aforementioned paper. The output
data consist of refractory wall temperature distribution, the axial solids and gas temperature profiles,
axial solids composition profile, the length required for drying of the solid charge and the total kiln length
required to achieve 98% conversion of anatase TiO2 to rutile TiO2. A detailed parametric study with
respect to the controlling parameters such as percent water content (with respect to dry solids), solids
flow rate, gas flow rate, kiln inclination angle and kiln rotational speed lent a good physical insight into
the rutile-TiO2 production process in a rotary kiln.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction chloride process using chlorine gas. The last process step of the sul-
phate method, named calcination is performed in rotary kiln and
This paper presents a computer model of heat transfer during has been considered in the present work.
production of titanium dioxide white pigment in rutile form in a
rotary kiln. 1.2. Description of rotary kiln

1.1. Production of rutile titanium dioxide (TiO2) in a rotary kiln
Titanium dioxide is a white solid inorganic substance which is
used as a pigment or whitener in paints, paper, plastics, textiles,
and other products. It occurs in several polymorphs, among them,
anatase and rutile are manufactured in the chemical industry as
white pigments. The pigment properties of rutile titanium dioxide
are better than that of anatase titanium dioxide and are of more
economical importance. Titanium dioxide white pigments are pro-
duced from a variety of ores by two different processes, namely,
the sulphate process using concentrated sulphuric acid and the
⇑ Corresponding author.
E-mail address: psg@iitk.ac.in (P.S. Ghoshdastidar).
A rotary kiln consists of a refractory lined cylindrical shell
mounted at a slight inclination from the horizontal plane (Fig. 1).
The kiln is rotated at a very low speed about its longitudinal axis
and the raw charge comprising hydrous titanium dioxide in a
moist cake form is fed into the upper end of the cylinder and a
hot combustion gas mixture at 1 bar flows from the other end.
The gas is a mixture of products on burning of natural gas in a sep-
arate combustion chamber.
In the present study, the kiln is considered to comprise three
sections. In the first section, the wet solids are heated to the satu-
ration temperature of water. In the second section, the liquid evap-
orates at constant temperature until the charge is completely
devoid of moisture. In the third section, the solids are heated till
the required degree of conversion of anatase to rutile titanium

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.10.024
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
264 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

Nomenclature

Ae;k frequency factor for kth reaction (s
1) qj net heat transfer for jth surface element in an axial seg-
Acg cross-section area of gas flow in an axial segment (m2) ment (wall or solids), (W)
Ag total surface area exposed to gas in an axial segment qr;z net heat transfer from gas to solids and wall (W)
(m2) R radius of the kiln, Fig. 2
ag gravitational acceleration (9.81 m/s2) Rex Reynolds number based on relative velocity between
Ai elemental area for ith element at inner wall in an axial wall and air outside the kiln
segment (m2) ReDh Reynolds number based on hydraulic diameter Dh
Aj,inner contact area between wall and solids per unit element Ru universal gas constant (8.314 J/mol K)
(m2) T g;z gas temperature at axial position z (K)
Cp specific heat at constant pressure (J/kg K) T j;inner inner wall temperature at jth surface element (K)
D diameter of the kiln (m) T s;z solids temperature at axial position z (K)
Dh hydraulic diameter of gas flow (m) Dt residence time (s)
dm_ k;z depletion rate of the reactant of kth reaction in an axial To,av average temperature at outer wall (K)
segment (kg/s) U circumferential speed of the kiln (m/s)
_ v;z
dm evaporation rate of moisture content in wet solid in vg mean velocity of gas (m/s)
concerned axial segment (kg/s) Vz axial velocity of solids (m/s)
Ea;k activation energy for kth reaction (J/mol) x radial coordinate (m), Fig. 2
Eb blackbody emission per unit area (W/m2) Xk mass fraction of the solid component k
f percentage of moisture content on dry basis in wet so- y circumferential coordinate (m), Fig. 2
lids z axial coordinate (m)
Fgj shape factor between gas and surface element j
Fij shape factor between surface elements i and j (including Greek letters
j = i) a fill angle (deg), Fig. 2
Gr Grashof number arf thermal diffusivity, Eq. (4)
hcgw average convective heat transfer coefficient from gas to b volumetric thermal expansion coefficient (K
1)
refractory wall in an axial segment (W/m2 K) C fill angle (radian)
hcj local convective heat transfer coefficient from gas to jth e emissivity
element of solids (W/m2 K) fk;z degree of conversion of kth reaction at axial position z
hfg latent heat of vaporization of water (J/kg) r Stefan–Boltzmann constant (5.67  10
8 W/m2 K4)
ho convective heat transfer coefficient from outer wall to h as defined in Fig. 2 (deg)
surroundings (W/m2 K) l dynamic viscosity (kg/m-s)
hws contact heat transfer coefficient between wall and solids m kinematic viscosity (m2/s)
(W/m2 K) n Darcy friction factor, used in Gnielinski [35] correlation
Dhi heat of reaction for ith reaction (J/kg) q density (kg/m3)
k thermal Conductivity (W/m K) sg transmissivity of the gas
L length of the kiln (m) Ds time step (s)
L1 length of the first section of the kiln (m) / kiln inclination angle (deg)
L2 length of the second section of the kiln (m)
L3 length of the third section of the kiln (m) Subscripts
m_ g;z gas mass flow rate at axial position z (kg/s) a air
m_ k;z mass flow rate of the kth component of solid charge at
cr chemical reaction
axial position z (kg/s) g gas
m_ s;z solids mass flow rate at axial position z (kg/s) gs gas to solids
Mwk molecular weight of kth component of solids (g/mol)
i element number of the wall or the solid
nk Order of kth reaction j element number of the wall or the solid
N total number of surface elements in an axial segment k number of reaction, also number of component in solid
Nk rotational speed of the kiln (rev/min) charge
Nr number of surface elements exposed to gas in an axial
l liquid
segment O2 oxygen
Ns number of surface elements covered by solid in an axial s solids
segment sh shell or outer wall
NuDh nusselt number based on hydraulic diameter Dh
SO2 sulphur dioxide
Pg wetted perimeter of gas flow in an axial segment (m) rf refractory wall
Pr Prandtl number v vapour
q1;z net heat transfer from gas and exposed wall to solids
w water
(W) ws wall to solids
q2;z net heat transfer from covered wall to solids (W) z at an axial distance z from the solids inlet
qcr;z net heat energy absorbed or released by chemical reac- z + Dz at a distance z + Dz from the solids inlet
tions at axial position z (W)
qgp;z thermal energy associated with the released gaseous
products of chemical reactions (W) Abbreviations
CFD Computational Fluid Dynamics
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
Fig. 1. Simplified schematic diagram of a rutile-TiO2 rotary kiln.
dioxide is achieved and then are released from the kiln. The calci-
nation process produces TiO2 pigment with a rutile content of
around 98%. The lengths of the first, second and third sections of
the kiln are denoted by L1, L2 and L3, respectively. The total length
of the kiln is L which is the sum of three individual lengths. Heat
transfer processes in the kiln, the nomenclature of the kiln and
the coordinate system are shown in Fig. 2.
1.3. Basic solution methodology
A model of heat transfer among the gas, refractory wall and
solid has been developed. The radiation heat transfer is modelled
by dividing the solid surface and the wall into surface elements.
The turbulent convection heat transfer from the gas to the wall
and solid is estimated by a stand-alone CFD model assuming a
non-rotating kiln with stationary solid bed. It is in this part that
FLUENT, a commercial software which applies the finite-volume
method is made use of. The local convective heat transfer coeffi-
cients obtained from the FLUENT simulation are then supplied to
the main computer program developed in-house. Since the refrac-
tory wall is alternately heated and cooled during each revolution,
the quasi-steady heat conduction in the wall is present. The energy
equation for the wall is discretized using the finite-difference tech-
nique. The mass and energy balance equations are solved for the
solid and the gas assuming only axial temperature variation, since
the solid and the gas are well-mixed in the cross-sectional plane of
the kiln, giving rise to uniform temperature in the transversal sec-
tion. The ‘‘well-mixed” assumption is valid since the solid bed
motion is in rolling mode (as the kiln speed is small) and the gas
flow is turbulent. A number of higher order as well as fractional
Fig. 2. Schematic cross-section of a rotary kiln showing heat transfer processes, the fill-angle and the coordinate system.
265
order chemical reactions occur. They are modelled by a modified
form of Arrhenius rate law.
1.4. Literature review
The literature on rotary kilns can be divided into two parts,
namely, (i) studies related to modelling without chemical reac-
tions; and (ii) studies related to modelling with chemical reactions.
The literature review that follows covers these aspects.
1.4.1. Studies related to modelling without chemical reactions
Sass [1] developed a computer model for heat transfer in a
rotary kiln dryer using empirical relationships for radiation heat
transfer calculations. Kamke and Wilson [2] developed a computer
model of a single-pass, rotary drum dryer for drying of wood par-
ticles with or without a center-fill flighting section. Cook and
Cundy [3] presented an analytical model for the heat transfer pro-
cess taking place between the heated wall of a rotating cylinder
and the solid contained in it with an initial moisture content equal
to 2–7% of the mass of the dry solids. Ghoshdastidar and Unni [4]
developed a steady-state heat transfer model of drying and pre-
heating of wet solids in the non-reacting zone of a cement rotary
kiln. Ghoshdastidar et al. [5] have reported simulation results
based on an improved heat transfer model for a rotary kiln used
for drying and preheating of wet iron ore. Ghoshdastidar and Agar-
wal [6] performed a numerical simulation and optimization study
of heat transfer in a rotary kiln used for drying and preheating of
wood chips with superheated steam. Liu et al. [7] obtained an ana-
lytical solution to predict axial solid transport in rotary kilns. Sch-
midt and Nikrityuk [8] carried out a 2D numerical simulation of
transient temperature distribution by granular mixing in a hori-
zontal rotary kiln. Sonavane and Specht [9] presented a finite-
element method based numerical analysis of heat transfer in the
wall of a rotary kiln.
1.4.2. Studies related to modelling with chemical reactions
This can further be sub-categorized into two parts, namely, (i)
production of titanium dioxide; and (ii) production of other
chemicals.
1.4.2.1. Production of titanium dioxide. Ginsberg and Modigell [10]
developed a comprehensive one-dimensional dynamic model of a
rotary kiln for the production of titanium dioxide white pigment.
266 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
Their numerical predictions matched very well with the measure-
ment data. Based on heat and mass transfer equations, Dumont
and Bélanger [11] described a simplified process model for making
of titanium dioxide in a rotary kiln. Roubal et al. [12] developed the
Kalman Filter design based on the reduced model of the process in
an attempt to control the titanium dioxide production in a rotary
kiln. In both studies, dehydration and desulphurisation steps were
not considered. Also, Koukkari et al. [13] introduced the RATEMIX
model by combining multicomponent thermodynamics with
chemical kinetics and applied in calcination of metatitanic acid
to produce rutile titanium dioxide in a rotary drum calciner.
1.4.2.2. Production of other chemicals. Spang [14] developed a
dynamic partial-differential equation model of cement kiln incor-
porating a flame model. Manitius et al. [15] described a mathemat-
ical model for production of aluminium oxide by calcination of
basic ammonium aluminium sulphate. A steady state mathemati-
cal model of cement kiln was presented by Guruz and Bac [16]
using zone method. Watkinson and Brimacombe [17] performed
experimental work on pilot-scale rotary kiln for calcination of
limestone with parametric study. Watkinson and Brimacombe
[18] observed the effects of enriching the combustion air with oxy-
gen in a rotary lime kiln. Hard and Mu [19] calculated sensitivities
of certain process variables on fuel rate and solids throughput in
modelling a phosphate nodulizing kiln. A detailed steady-state
heat transfer model of burning of Plexiglas in a rotary kiln was
developed by Ghoshdastidar et al. [20]. A 3D steady state model
of a rotary calcining kiln was presented by Bui et al. [21] for the
petroleum coke. Davis [22] reported a model for magnetite oxida-
tion during iron ore pellet induration in a rotary kiln. A model was
developed for iron oxide pelletizing to simulate effects of under
bed injection on kiln fuel requirements and magnetite oxidation
by Davis [23]. Georgallis et al. [24] presented a 3D model for rotary
lime kiln which included evolution and combustion of species and
granular bed motion with calcination reaction. Marias et al. [25]
modelled pyrolysis of aluminium waste in a rotary kiln. Mintus
et al. [26] predicted solid composition and temperature profiles
and fuel requirement based on their one-dimensional cell model
for wet process rotary cement kilns. Mujumdar and Ranade [27]
presented a 1D model to analyse key processes occurring in solid
bed of cement kilns. Mujumdar et al. [28] developed Rotary
Cement Kiln Simulator by integrating the separate models for
pre-heater, calciner, rotary kiln and cooler. Flow and transport pro-
cesses were modelled in a calciner for cement production by
Fidaros et al. [29]. Shahriari and Tarasiewicz [30] modelled a clin-
ker rotary kiln using operating functions concept.
1.5. Objectives of the present work
The above literature review reveals that only a few models [10–
13] of titanium dioxide rotary kiln are available. The main differ-
ence between the present and aforementioned ones is that the lat-
ter do not have the capability of predicting the kiln length required
to produce rutileTiO2 for a given set of input parameters. The pre-
sent work is an attempt to fill this gap in the existing literature.
The objectives of the present study are as follows.
1. To develop a detailed computer model of heat transfer with
chemical reactions during the production of rutile titanium
dioxide in a rotary kiln;
2. To compare the total predicted kiln length required to achieve
the desired solids composition as predicted by the present
model with that of the actual kiln reported in Ginsberg and
Modigell [10];
3. To predict axial distributions of solid temperature, gas temper-
ature, composition of the solid charge within the kiln;
4. To conduct a detailed parametric study with respect to mois-
ture percent (dry basis), mass flow rate of the solid, mass flow
rate of the gas, kiln inclination angle and kiln rotational speed
in order to get a good physical insight into the manufacturing
process of rutile titanium dioxide in a rotary kiln.
2. Problem formulation
In this section the reaction processes, modelling of radiation
and conduction, gas convection, reaction kinetics, and, mass and
energy balances in the solids and gas are presented in detail.
2.1. Reaction Processes Occurring in a Rutile-TiO2 Rotary Kiln
In the sulphate process, ilmenite or titanium slag as raw mate-
rial is digested with sulphuric acid and results in a solution con-
taining TiOSO4 and FeSO4 . After removing iron and other
impurities by vacuum crystallization from the solution, titanium
sulphate is hydrolysed to hydrous titanium dioxide. It largely con-
sists of slurry of TiOðOHÞ2 (metatitanic acid) and a relatively small
amount of TiOSO4  H2 O. This solid charge is then fed into a rotary
kiln to undergo calcination process and finally, rutile titanium
dioxide is produced. In addition to moisture evaporation at satura-
tion temperature, following chemical reactions take place inside
the rotary kiln as wet solids move from the inlet to the outlet of
the kiln.
Dehydration of metatitanic acid and TiOSO4  H2 Ooccurs in the
range of 100–250 °C [31] (Reactions 1 and 2). Reaction 3 results
in loss of sulphur trioxide (which immediately decomposes into
SO2 and O2) from titanyl sulphate at about 650 °C [32] and finally,
phase transformation of anatase to rutile takes place exothermally
in the range of 700–950 °C [32] (Reaction 4).
The above mentioned reactions are listed in Table 1a along with
their respective heat of reactions, pre-exponential factors, activa-
tion energies and orders of reaction. The composition of wet solids
at the solid inlet is given in Table 1b.
2.2. Radiation exchange among hot gas, refractory wall and the solid
surface
Since the gas temperature is high, thermal radiation plays an
important role. The radiation is modelled by dividing the wall into
surface elements as shown in Fig. 3. Each axial segment of the
refractory surface is divided into Nr surface elements of equal size.
The solid surface is divided into Ns surface elements. This can be
considered as a 2-D enclosure since the surface elements are quite
long. The temperature of the solid surface element and the gas are
assumed to be uniform in each axial segment. The wall surface ele-
ments are assumed diffuse and gray. It is to be kept in mind that
the surface elements exchange radiation only with the surfaces
of the same axial segment which is sufficiently long. Since the
hot gas contains CO2 and H2O and it is treated as radiatively
participating.
The radiation heat transfer is computed by using the theory of
Hottel [33] for a gray enclosure containing a gas which emits,
absorbs and transmits radiation. If the gas volume is enclosed by
N gray surface elements at different temperatures, the net energy
gain, qj at a particular surface can be expressed as
" #
eþ1 X
N
qj ¼ Ag F gj eg Eb;g þ Ai F ij sg Eb;i
Eb;j Aj ð1Þ
2 i¼1
where e is the emissivity of the wall. Eq. (1) is valid when e is
greater than 0.8, which is true in the present case. For the calcula-
tion of shape factors, F ij expressions derived by Ghoshdastidar
et al. [20] have been used. eg and sg are the emissivity and transmis-
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279 267

Table 1a
Reactions, kinetics and heat of reaction data (Ginsberg and Modigell [10]).

S. No. Reaction Equation Frequency Factor Activation Energy Heat of Reaction Order of Reaction
Ae;k (s
1) Ea;k (kJ/mol) Dhk (kJ/mol) nk

1 TiOðOHÞ2 ! TiO2 ðaÞ þ H2 OðgÞ 4.1  103 45 90 3.00

2 TiOSO4  H2 O ! TiOSO4 þ H2 OðgÞ 5.0  105 75 90 0.50
3 TiOSO4 ! TiO2 ðaÞ þ SO2 ðgÞ þ 0:5O2 5.0  109 230 396 0.30
4 TiO2 ðaÞ ! TiO2 ðrÞ 1.7  1023 491 -5.6 0.67

Table 1b
Mass fractions of reactants (Ginsberg and Modigell [10]).

S. No. Reactant Mass fraction at solid inlet (w.r.t dry solids)

1 TiOðOHÞ2 0.8894
2 TiOSO4  H2 O 0.1106

Fig. 3. Section of the kiln showing surface elements of the wall and the solid.

sivity of the gas. The value of gas emissivity has been taken directly
from Ginsberg and Modigell [10]. The gas transmissivity has been
calculated using the expression,sg ¼ 1
eg , assuming gas reflectiv- Fig. 4. A two-dimensional grid showing triangular elements in a cross-sectional
plane of the TiO2 kiln.
ity to be zero and Kirchhoff’s law of radiation to be valid.

2.3. Gas convection

2
In direct fired kilns of small diameter operating at solids tem- n ¼ ð0:79lnðReDh Þ
1:64Þ ð2aÞ
perature up to 1100 K gas convection may be significant (Brima-
Gnielinski correlation is valid for,
combe and Watkinson [34]). In order to bring more precision in
the calculation of connective heat transfer from main heat source 0:5 6 Pr 6 2000 and 3000 6 ReDh 6 5  106
in the rotary kiln, that is hot gas mixture, local convective heat
transfer coefficients on the exposed surfaces of refractory and solid NuDh and ReDh are the Nusselt number and Reynolds number,
bed are estimated a priori, via a secondary model of heat transfer. respectively, based on hydraulic diameter for the gas region, Dh .
In this secondary model, a non-rotating kiln containing a station- Dh is expressed as
ary solid bed is simulated using a finite volume-based commercial 4Acg
CFD package ANSYS-FLUENT 15.0. An unstructured mesh with tri- Dh ¼ ð2bÞ
Pg
angular cross-section prisms as elements is generated using com-
mercial software Gambit 2.4.6. Fig. 4 shows a two-dimensional where Acg and Pg are cross-sectional area and wetted perimeter of
grid constructed via triangular elements in a cross-sectional plane gas flow in an axial segment, respectively.
of the TiO2-kiln.  
pD2 C sin C
The length of the kiln used for this simulation is predicted from Acg ¼ 1
þ ð2cÞ
4 2p 2p
the main model. Since local convective heat transfer coefficients
are not known in advance, average values for gas–solids and gas-   a
360
a
refractory contact surfaces at each axial location are estimated P g ¼ pD þ D sin ð2dÞ
360 2
using a correlation (Eq. (2)) for Nusselt number developed by
Gnielinski [35] for turbulent flow in a smooth pipe. where C and a are the fill angle of solids in radians and degrees,
respectively.
ðn=8ÞðReDh
1000ÞPr
NuDh ¼ ð2Þ ReDh is expressed as
1 þ 12:7ðn=8Þ1=2 ðPr2=3
1Þ
qg v g Dh 4m_ g
where, n is the Darcy friction factor and for smooth tubes, which is ReDh ¼ ¼ ð2eÞ
lg P g lg
given by
268 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

where qg , v g , lg and m
_ g are the average density, mean velocity, covered by solids exchange heat with solid through surface contact
dynamic viscosity and mass flow rate of gas mixture, respectively. (Helmrich and Schugërl [39]). It may be noted that the sticking of
The average convective heat transfer coefficient from gas to some rutile TiO2 to the inner refractory wall at various axial loca-
refractory wall in an axial segment is given by tions will alter the contact heat transfer coefficient which can only
be obtained from an experiment on a similar kiln. The authors
NuDh kg don’t have any data regarding this aspect available in the published
hcgw ¼ ð2fÞ
Dh literature and hence, an average value of contact heat transfer coef-
where kg is the mean thermal conductivity of the gas mixture. The ficient (377 W/m2 K) for the TiO2 kiln as estimated by Ginsberg and
average convective heat transfer coefficient from gas to solids in an Modigell [10] has been used.
axial segment is assumed to be an order of magnitude higher than The heat transfer coefficient for convection from outer wall to
that from gas to refractory wall [36,37]. surroundings is calculated as (Suryanarayana et al. [38]):
The Reynolds Number based on the hydraulic diameter of gas " ! #0:35
0:11ka Re2x
flow cross-section is found to be 101,672. The standard k-omega ho ¼ þ Gr Pra ð4Þ
turbulence model is implemented. The conjugate heat transfer is
Dsh 2
considered at the interfaces of gas–solids and gas-refractory.
where, Dsh is the outer diameter of the kiln, ka and Pra are the ther-
Thermo-physical properties of gas mixture, refractory wall and
mal conductivity and Prandtl number of the surrounding air respec-
bulk solid are listed in Table 2. At the solid inlet of the kiln the mass
tively, Rex is the Reynolds number based on relative velocity
flow rate and temperature of gas mixture are as given in Table 2. At
between the cylindrical wall and the external air surrounding the
the outer peripheral wall and end walls of the kiln, a mixed bound-
kiln. Gr is the conventional Grashof number characterizing natural
ary condition is specified with a heat transfer coefficient as esti-
convection.
mated using a correlation given by Suryanarayana et al. [38]. The
momentum, turbulent kinetic energy, turbulent dissipation rate pD2sh x
Rex ¼ ð4aÞ
and energy equation are solved with the second order upwind dif- 2ma
ferencing scheme. The convergence criteria selected for the gov-
where, x is the angular velocity of the rotary kiln and ma is the kine-
erning equations are 10
3 for the continuity, momentum, k and
matic viscosity of surrounding air.and
omega equations, and 10
6 for the energy equation.
A grid independence test is performed based on average wall ag bðT o;av
T a ÞD3sh
Gr ¼ ð4bÞ
function heat transfer coefficients at the interfaces of gas–solids m2a
and gas-refractory. The optimal value of total number of cells used
in the 3D computational domain is 3154424. The computations are where, ag is the gravitational acceleration, b is the volumetric ther-
performed on SONY workstation with core i5 processor and 4 GB mal expansion coefficient, T o;av is the average temperature on outer
memory by using ANSYS-FLUENT 15.0. wall of the kiln and T a is the surrounding temperature outside the
The local convective heat transfer coefficients for each circum- kiln.
ferential element (Fig. 3) in each axial segment are obtained from
the aforesaid simulation. The range of the same is found to be 2.5. Reaction rates
12.16 W/m2K to 21.08 W/m2K. These local values are imported
into the main model (at the same locations in computational For any reaction, the rate of change in degree of conversion of
domain of the main model) as input data and convection heat the reaction is assumed to be governed by kinetic expression (Eq.
transfer from gas to solids and gas to refractory is estimated. (5)), which is a modified form of Arrhenius rate law:
 
dfk;z Ea;k
2.4. Conduction in the refractory wall ¼ Ae;k exp
ð1
fk;z Þnk ð5Þ
dt Ru T s;z

The conduction in the refractory wall is modelled assuming where fk;z denotes the degree of conversion of kth reaction at axial
quasi-steady heat conduction since wall elements of the kiln are position z, Ru is the universal gas constant, T s;z is the mean solids
alternately heated and cooled during each revolution by gas and temperature at axial position z, Ae;k , Ea;k , nk are the frequency factor,
solid, respectively. Only radial heat conduction in the wall is taken activation energy and order of kth reaction, respectively. The values
into account, assuming negligible conduction in the circumferen- of Ae;k , Ea;k and nk are provided in Table 1a.
tial and axial direction as the kiln dimensions are much larger in The degree of conversion, fk;z is defined as
those directions. A non-rotating coordinate system is used to _ k;0
m
m _ k;z
model the wall conduction equation (Eq. (4)). Because the refrac- fk;z ¼ ð6Þ
_ k;0
m
m _ k;f
tory thickness (0.25 m) is appreciably smaller than kiln diameter
(2.3 m), the ratio being about 10.9%, the effect of curvature is
neglected and hence, Cartesian coordinates are used instead of
cylindrical coordinates for the sake of simplicity.

@T @2T
U ¼ arf 2 ð3Þ Table 1c
@y @x Molecular masses of various components in solid charge.

where U is the circumferential speed of the wall and arf is the ther- S. No. Component Molecular weight
k Mw (g/mol)
mal diffusivity of the wall.
The boundary condition for the inner wall depends on whether 1 TiOðOHÞ2 97.88
2 TiOSO4  H2 O 177.94
the surface element is exposed to the gas or it is covered by solids.
3 TiOSO4 159.93
The radiation heat transfer, qj from the gas to the inner surface ele- 4 TiO2 ðaÞ 79.87
ments is obtained from Eq. (1). In addition, convection heat trans- 5 TiO2 ðrÞ 79.87
fer is calculated from the local convective heat transfer coefficients 6 H2 OðlÞ 18.02
7 H2 OðgÞ 18.02
as discussed in Section 2.3. Surface elements of the kiln that are
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279 269

where m _ k;0 and m _ k;f are the initial and final mass flow rates of the ichiometry of reactions 1, 3 and 4, its new mass flow rate at exit
reactant of kth reaction respectively, m _ k;z is its mass flow rate at of this axial segment is given as,
any axial position z.    
Mw4 Mw4
With the help of degrees of conversion of the reaction at any _ 4;z exit ¼ ½m
½m _ 4;z entry
dm
_ 4;z þ _ 1;z þ
dm _ 3;z
dm ð10Þ
Mw1 Mw3
axial segment, using Eq. (6), the amount of depletion per second
for different species of solid charge ðdm _ k;z ) are estimated. Using where ½m_ 4;z entry is the mass flow rate of TiO2 ðaÞ at entry of axial seg-
this information, based on stoichiometry of the chemical reactions, ment and dm _ 4;z is the decomposed amount during this reaction in
updated mass flow rates of solid species (that is, the mass flow that axial segment. Mw4 is the molecular weight of TiO2 ðaÞ.
rates at the exit of that axial segment, m _ k;z ) are estimated. Mass flow rate of TiO2 ðrÞ, as it is generated from reaction 4, is
Table 1c lists the components in solid charge with corresponding updated as well in a similar way,
index k and their respective molecular weights.
_ 5;z exit ¼ ½m
½m _ 5;z entry þ dm
_ 4;z ð11Þ
_ k;z entry and ½m
If ½m _ k;z exit are the mass flow rates of kth species at
the entry and exit of axial segment ‘z’ and dm _ k;z is the amount of _ 5;z entry and ½m
where, ½m _ 5;z exit are the mass flow rates of TiO2 ðrÞ at
depletion per second of reactant in kth reaction in that segment, entry and exit of that axial segment.
then using molecular masses Mwk , following expressions are Mass flow rate of water vapour that is released as gaseous pro-
obtained for updated mass flow rates at the exit of the segment. duct from reaction 1 and 2, is tracked at each axial segment before
Eqs. (7)–(12) show the expressions for updated mass flow rates it is mixed into hot gas mixture. Again based on stoichiometry of
of solid components in each axial segment and are written based reactions 1 and 2, amount decomposed per second in segment is
on the stoichiometry of the chemical reactions. estimated as,
Reaction 1:    
Mw7 Mw7
_ 7;z ¼
dm _ 1;z þ
dm _ 2;z
dm ð12Þ
TiOðOHÞ2 ! TiO2 ðaÞ þ H2 OðgÞ Mw1 Mw2
In reaction 1, reactant TiOðOHÞ2 or metatitanic acid is thermally Here, Mw7 is the molecular weight of water.
decomposed into TiO2 (a) and water vapour. TiOðOHÞ2 is denoted From the knowledge of enthalpies of reactions, heat generation
by index 1, as shown in table. Updated mass flow rate of this com- due to chemical reactions in an axial segment can be estimated and
pound at exit of axial segment is given by following expression. expressed as:
_ 1;z exit ¼ ½m
½m _ 1;z entry
dm
_ 1;z ð7Þ _ 1;z Dh1 þ dm
qcr;z ¼ dm _ 2;z Dh2 þ dm
_ 3;z Dh3 þ dm
_ 4;z Dh4 ð13Þ

where ½m_ 1;z entry is the mass flow rate of TiOðOHÞ2 at entry to the where, Dh1 , Dh2 , Dh3 and Dh4 are the heat of reactions for Reactions
segment and dm _ 1;z is the amount of depletion of TiOðOHÞ2 per sec- 1 to 4 and are given in Table 1a.
ond in that segment.
Reaction 2: 2.6. Mass and energy balances in solids and gas

TiOSO4  H2 O ! TiOSO4 þ H2 OðgÞ As mentioned earlier, the kiln is considered to be consisting of

Similarly, in reaction 2, reactant TiOSO4  H2 O is decomposed into three sections. In the first section, wet solid charge is heated to
TiOSO4 and water vapour. TiOSO4  H2 O is denoted by index 2, as the saturation temperature of the entrained liquid. In the second
shown in table. Its new mass flow rate at exit of axial segment is section, liquid evaporates at constant temperature and in the third
given by following expression. section dry solid charge is heated till the required conversion of
anatase TiO2 to rutile TiO2 is achieved.
_ 2;z exit ¼ ½m
½m _ 2;z entry
dm
_ 2;z ð8Þ The mass and energy balance of the solid in each segment either
in first and third segment gives the expression for exit solid tem-
where ½m_ 2;z entry is the mass flow rate of TiOSO4  H2 O at entry to seg-
perature of that segment, T s;zþDz for that section while the same
ment and dm _ 2;z is its decomposed amount per second during that performed on an axial segment in second section of the kiln gives
axial segment. _ v;z . The end of the sec-
the expression for the rate of evaporation, dm
Reaction 3: ond section is indicated where the cumulative m _ v is equal to the
TiOSO4 ! TiO2 ðaÞ þ SO2 ðgÞ þ 0:5O2 predetermined amount of water to be evaporated per second. Sim-
ilarly, mass and energy balance of hot gas contained in an axial
TiOSO4 is a product of reaction 3 and is used as a reactant in reac- segment in any of the three sections of the kiln gives the expres-
tion 3. TiOSO4 is denoted by index 3, as shown in table. Based on sion for T g;zþDz for that section.
stoichiometry of both reactions, following expression can be writ-
ten for updated mass flow rate of TiOSO4 at exit to that axial 2.6.1. Mass balance in Sections I and III
segment. In solids region (Fig. 5),
 
Mw3 m _ s;z
dm
_ s;zþDz ¼ m _ g;z ð14Þ
_ 3;z exit ¼ ½m
½m _ 3;z entry
dm
_ 3;z þ _ 2;z
dm ð9Þ
Mw2
_ 3;z entry is the mass flow rate of TiOSO4 at entry to axial seg-
where ½m
ment and dm _ 3;z denotes the amount depleted per second in that
segment. Mw2 and Mw3 are the molecular weights of TiOSO4  H2 O
and TiOSO4 respectively.
Reaction 4:
TiO2 ðaÞ ! TiO2 ðrÞ
TiO2 ðaÞ is produced from reactions 1 and 3. It is consumed in reac-
tion 4, where it is converted into final product that is TiO2 ðrÞ.
TiO2 ðaÞ is denoted by index 4, as shown in table. According to sto- Fig. 5. Mass balance of any axial segment in first and third sections of the kiln.
270 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

where m _ s;z and m

_ s;zþDz are the mass flow rates of the solids at any Dz
Dt ¼ ð23Þ
axial position z and z + Dz respectively. dm _ g;z is the total amount Vz
of released gaseous products from chemical reactions in an axial
where V z is the axial velocity of the solids. Note that V z ¼ U tan /.
segment.
For derivation of this expression see Ghoshdastidar and Agarwal [6].
_ g:z ¼ dm
dm _ H2 OðgÞ;z þ dm
_ SO2 ðgÞ;z þ dm
_ O2 ðgÞ;z ð15Þ So,

Here, dm _ H2 OðgÞ;z , dm
_ SO2 ðgÞ;z and dm
_ O2 ðgÞ;z are the amounts of H2 OðgÞ, X
Ns X
Ns
q1;z ¼ qj þ hcj Aj ðT g;z
T s;z Þ ð24Þ
SO2 ðgÞ and O2 ðgÞ releasing from chemical reactions in solids in that
j¼1 j¼1
axial segment.
In gas region (Fig. 5), where qj is the total radiation heat transfer to the jth element of
m _ g;z
dm
_ g;zþDz ¼ m _ g;z ð16Þ solid.

_ g;z and m
_ g;zþDz are the mass flow rates of the gas at any axial X
Ns
where m q2;z ¼ hws Aj;inner ðT j;inner
T s;z Þ ð25Þ
position z and z + Dz respectively. and, j¼1

_ w;s ¼ m
m _ w;g ð17Þ _ H2 OðgÞ;z cp;v þ dm
qgp ¼ T s;z ðdm _ SO2 ðgÞ;z cp;SO2 þ dm
_ O2 ðgÞ;z cp;O2 Þ ð26Þ
where m _ w;g is the mass flow rate of evaporated moisture content in In gas region (Fig. 7),
first section, that is accumulated in gas from the second section of
_ g;z cp;g þ m
T g;z ðm _ w;g cp;v Þ þ qr;z
qgp;z
the kiln. T g;zþDz ¼ ð27Þ
m_ g;zþDz cp;g þ m _ w;g cp;v
2.6.2. Mass balance in Section II where qr;z is the total heat transfer from gas to solids and wall. cp;g is
In solids region (Fig. 6), the specific heat of gas at constant pressure.
m _ s;z
dm
_ s;zþDz ¼ m _ g;z ð18Þ
2.6.4. Energy balance in Section II
and,
In solids region (Fig. 8),
m _ w;s;z
dm
_ w;s;zþDz ¼ m _ v;z ð19Þ q1;z þ q2;z
qgp;z
qcr;z
_ v;z ¼
dm ð28Þ
where, m_ w;s;z is the mass flow rate of moisture content in wet solids hfg
in second section of the kiln at any axial position z.
where, dm _ v;z is the evaporation rate of moisture content in dry solid
In gas region (Fig. 6),
in concerned axial segment and hfg is the latent heat of vaporization
m _ g;z
dm
_ g;zþDz ¼ m _ g;z ð20Þ of water in J/kg.
For this section,
and,
_ w;g;zþDz ¼ m
m _ v;z
_ w;g;z
dm ð21Þ T s;zþDz ¼ T s;z ¼ 373 K
¼ Saturation temperature of water at 1 bar
where m_ w;g;z is the mass flow rate of evaporated moisture content
present in gas in second section of the kiln at any axial position z. In gas region (Fig. 9),
_ g;z cp;g þ m
T g;z ðm _ w;g;z cp;v Þ
T s;z dm
_ v;z cp;v þ qr;z
qgp;z
2.6.3. Energy balance in Section I T g;zþDz ¼ ð29Þ
In solids region (Fig. 7), _ g;zþDz cp;g þ m
m _ w;g;zþDz cp;v

_ s;z cp;s þ m
T s;z ðm _ w;s cp;l Þ þ q1;z þ q2;z
qgp;z
qcr;z
T s;zþDz ¼ ð22Þ 2.6.5. Energy balance in Section III
_ s;zþDz cp;s þ m
m _ w;s cp;l
In solids region (Fig. 10),
where, m _ w;s is the mass flow rate of moisture content present in wet
_ s;z C p;s þ q1;z þ q2;z
qgp;z
qcr;z
T s;z m
solids in first section of the kiln. It remains same throughout the T s;zþDz ¼ ð30Þ
section. cps and cpl refer to specific heats (at constant pressure) for _ s;zþDz cp;s
m
dry solids and moisture content respectively. Also, q1;z is the heat In gas region (Fig. 10),
transfer to the solids from gas and exposed surface elements of
the kiln through convection and radiation, q2;z is the heat trans- _ g;z cp;g þ qr;z
qgp;z
T g;z m
T g;zþDz ¼ ð31Þ
ferred to the solid through surface elements of kiln in direct contact m _ g;zþDz cpg
with solids, qgp;z is the thermal energy associated with the released
gaseous products of chemical reactions and qcr;z is the heat gener-
3. Overall method of solution
ated by the chemical reactions in Dz distance or Dt time, where
Dt is computed from
A computer program in FORTRAN 95 was developed to obtain
the numerical results for the present problem. The input data
required for the program are shown in Table 2. The specific heat
of the dry solids is calculated by using Eq. (34) from the specific
heat of individual components listed in Table 1c.
X
5
C p;s ¼ X k C p;k ð32Þ
k¼1

where X k and C p;k are the mass fraction and specific heat of compo-
nent k respectively. k = 1 to 5 correspond to TiOðOHÞ2 , TiOSO4  H2 O,
Fig. 6. Mass balance of any axial segment in second section of the kiln. TiOSO4 , TiO2 ðaÞ and TiO2 ðrÞ respectively.
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
Fig. 7. Energy balance of any axial segment in first section of the kiln.
Fig. 8. Energy balance for solids in any axial segment in second section of the kiln.
A mixture density relation is used to calculate the effective den-
sity of wet solids from the densities of the components as given in
Table 1d.
qs;dry solids þ 100
f
ql
qs ¼ ð33Þ
1 þ 100
f
where f is the percent of moisture content with respect to dry
solids, ql is the density of liquid water and qs;dry solids is the density
of dry solids.
1 sensitivity analysis for axial segment length, Dz with respect to
qs;dry solids ¼ ð34Þ
X
5 predicted kiln length is also performed. It is found that the pre-
X k =qs;k
k¼1
where qs;k is the density of the component k in solid charge.
The solution is initiated at the solids entrance of the kiln and
proceeds to the outlet. In an axial segment, the gas and solids tem-
perature being known, the steady state refractory wall tempera-
ture distribution is calculated by solving Eq. (3) through the False
Transient Approach. The solid temperature and mass flow rate
are known at the inlet. Similarly, gas temperature and gas mass
flow rate are known at the solid inlet (which is actually the exit
for the gas since the gas is flowing countercurrent to the solid
flow). Thus, the energy balance equations (see Section 2.6) are
solved for the first segment, and solid composition, solids, and
gas temperature are obtained at the exit of the first section. Once
Fig. 9. Energy balance for gas in any axial segment in second section of the kiln.
271
the new composition of the solids at the inlet of the second seg-
ment is determined, the new cross-sectional area of the solids
and the new fill-angle are estimated and hence, new shape factors
are computed. The solids and gas temperature, and species compo-
sition at the exit of the second segment can be calculated in the
same way as described earlier. The solution then proceeds by the
analysis of each succeeding segment. The process is continued till
the end of the kiln. The end of the kiln is indicated by the position
where the degree of conversion for the last reaction, that is, ana-
tase to rutile titanium dioxide conversion has reached 98%. Thus
the kiln length can be computed.
A time step of Ds = 0.1 s is used for calculations. Since the initial
temperature distribution of the refractory wall is not known, an
arbitrary temperature is assumed at all grid points in the refractory
wall. The solution converges where there is no further change in
the temperature at each grid point as s ? 1. This temperature dis-
tribution represents the steady state temperature distribution of
the wall.
Grid independence tests have been performed to obtain the
optimum grid spacing values in the circumferential and radial
directions. The number of grid points used in the circumferential
direction is 150 and that in the radial direction is 61. In addition,
dicted kiln length and hence axial solid and gas temperature distri-
butions remain more or less unchanged while Dz is changed from
0.10 to 0.25 m. The largest value of Dz in this range is taken in
order to save CPU time. Therefore, Dz used in this study is 0.25 m.
The simulations are performed on a high performance comput-
ing system (having Terra-flops rating). The CPU time required for a
simulation with the input data (Table 2) is approximately 46.43 h.
The output data consist of refractory wall temperature, solids tem-
perature, gas temperature, solids composition, individual lengths
of first, second and third sections of the kiln (L1, L2, L3), and the
total kiln length (L). The capability of this model to predict the kiln
length is unique and has not been found in earlier TiO2-kiln mod-
els. The overall solution algorithm is described next in a step-by-
step fashion.
272 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

Fig. 10. Energy balance of any axial segment in second section of the kiln.

Overall Solution Algorithm Table 1d

Material properties of the solid components (Ginsberg and Modigell [10]).
1. Input the data for various kiln geometrical, flow, thermo-
S.No. Component Density (kg/m3) Specific heat (J/mol K)
physical and chemical kinetics parameters.
2. Input the solids composition, solids temperature and gas 1 TiOðOHÞ2 2540 142
2 TiOSO4  H2 O 2430 208
temperature at solids inlet.
3 TiOSO4 2890 133
3. Set m =1 for the first axial segment. 4 TiO2 ðaÞ 3900 70
4. Calculation of solid fill angle in the segment. 5 TiO2 ðrÞ 4240 70
5. Shape factor calculations for radiative heat echange among 6 H2 OðlÞ 1000 76
gas, wall and solid bed.
6. Calculation of heat transfer by convection from Eq. (2) and
radiation from Eq. (1). Table 2
7. Estimation of refractory wall temperature distribution by Input data to the program (Ginsberg and Modigell [10]).
solving energy equation by false transient method using 1. Rotary Kiln
explicit finite-difference scheme. (a) Diameter (inner) 2.3 m
8. Calculation of total heat transfer (convection + radiation) (b) Refractory
from gas and to solids in that segment. (i) Thickness 0.25 m
(ii) Thermal Conductivity 1.6 W/m K
9. Calculation of mass flow rates of solid species at the exit of (iii) Specific heat 950 J/kg K
that segment using modified Arrhenius equation for each (iv) Density 2310 kg/m3
reaction. (v) Emissivity 0.8
10. Calculation of thermal energy absorbed or released in that (c) Rotational speed 0.33 rpm
(d) Angle of inclination 2.29°
segment due to chemical reactions in the solid bed.
2. Solid
11. Calculation of the temperatures of solids and gas at the (a) Inlet temperature 308 K
exit of that segment using global mass and energy balances (b) Mass flow rate (dry) 1.309 kg/s
for solids and gas, respectively. (c) Percent water (on dry basis) 75.14
12. Update the gas and solids flow rates at the exit of the (d) Emissivity 0.8
3. Gas
segment. (a) Outlet temperature 648 K
13. Check whether degree of conversion of final reaction (i.e., (b) Specific heat 1356 J/kg K
that is conversion of anatase titanium dioxide to rutile (c) Mass flow rate 6.218 kg/s
titanium dioxide) has reached 98%. If it has reached this (d) Emissivity 0.422
(e) Dynamic viscosity 3.58  10
5 kg/m-s
value, print the total kiln length, wall temperatures and
4. Water Vapor
axial distributions of solids composition, solids (a) Latent heat of vaporization$ 2.258  106 J/kg
temperature and gas temperature. (b) Specific heat $ 2042.71 J/kg K
14. If it has not reached, then proceed to next axial segment, 5. Specific heat of sulfur dioxide$ 627.99 J/kgK
m=m+1, and repeat the procedure from steps 4 to 13. 6. Specific heat of oxygen$ 919.83 J/kgK
7. Contact heat transfer coefficient 377 W/m2K
15. Grid independence test in two-dimensional refractory 8. Ambient temperature outside the kiln 287.15 K
zone and sensitivity analysis for axial segment length with
$
respect to predicted kiln length are performed. Chase [40].

16. Optimum kiln length obtained in step 15 is used in the

secondary model in ANSYS-FLUENT (as discussed in
Section 2.3).
17. From the secondary model, local convective heat transfer
coefficients from gas to wall and solids are obtained and are
imported back to the main model at the same physical
locations.
18. Same procedure consisting of steps 1 to 14 are repeated in
this modified model and final output is received in terms of
total kiln length, wall temperatures and axial distributions
of solids composition, solids temperature and gas
temperature.
4. Results and discussion
This section firstly presents comparison of the results with that
of an earlier work. Following that, results of the parametric study
are shown to see the effect of variation in moisture content (dry
basis), dry solids mass flow rate, gas mass flow rate, rotational
speed of the kiln, and angle of inclination of the kiln on the axial
solid and gas temperature distributions, axial profiles of percent
conversion of anatase TiO2 to rutile TiO2 as well as on the predicted
length of the kiln and the rate of production of rutile TiO2.
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
4.1. Validation of Results with the Numerical Results of Ginsberg and
Modigell [10]
The kiln length predicted by present model turns out to be
45.75 m as compared to the actual kiln length of 45 m reported
in Ginsberg and Modigell [10]. The predicted kiln length is only
1.67% larger than the actual kiln, indicating an excellent match.
Fig. 11 illustrates the comparison of the present axial solid and
gas temperature profiles with those computationally obtained by
Ginsberg and Modigell [10]. It may be noted that no steady-state
experimental data are found in Ginsberg and Modigell [10]. It
can be seen from Fig. 11 that general trends of both the profiles
show a good agreement with Ginsberg and Modigell [10]. In the
first section of the kiln, solid temperature increases slowly to the
saturation temperature of the wet solids and it takes about 37%
of the kiln length. At this point, the second section begins and
evaporation of water takes place. The temperature remains con-
stant at the saturation point. The second section covers about
47% of the kiln length and is the largest among all the three sec-
tions. The length of this section largely depends on the amount
of moisture present in the wet solids. Right from the beginning
of the third section, temperature increases at a higher rate due to
large heat transfer from hotter gas and wall in that region. From
about 92% to 98% of the kiln length, there is decay in @T @z
s
. It is caused
by higher energy requirement of third reaction, where decomposi-
tion of titanyl sulfate occurs. The temperature increases thereafter
till the final reaction, that is, the phase transition of TiO2 (a) to TiO2
(r) is completed. The final reaction is exothermic in nature and
hence, unlike other reactions, it releases thermal energy to gas
region. Completion of this reaction also marks the end of the third
section of the kiln length. The predicted temperature profiles also
qualitatively agree with Dumont and Bélanger [11] which shows
that pre-heating and moisture evaporation take place in nearly
70% of the kiln length and solid temperature rapidly rises there-
Fig. 11. Axial solid and gas temperature distributions: Comparison with the numerical solution of Ginsberg and Modigell [10].
273
after towards solid outlet similar to that predicted by the present
computer model.
4.2. Axial composition profile of the solid charge
Fig. 12 shows axial composition distribution moisture content,
reactants and the final product. The graph clearly shows the pre-
sent model has been able to simulate the chemical reactions prop-
erly. The thermal decomposition of TiOðOHÞ2 and TiOSO4  H2 O
(that is, Reactions 1 and 2 as shown in Table 1a) starts close to
the kiln entrance. In the first section of the kiln, Reaction 1 pro-
ceeds at much faster rate than Reaction 2. It starts slowing down
at the beginning of the second section of the kiln and remains slow
till any moisture is left in solid charge. This is because of the fact
that solid temperature remains constant within second section.
In the third section, due to increase in solid temperature, Reaction
1 speeds up for a while and then remains slow and steady. Reaction
2 also progresses rapidly in the third section. The production of
titanyl sulfate from this reaction and increasing solid temperature
lead to commencement of Reaction 3 where the desulphurization
produces TiO2(a) and SO3. The SO3 that is released is immediately
decomposed into SO2 and 0.5O2. This reaction is over at 954 K.
Before completion of Reactions 3 and 1 (Table 1a), final reaction
of phase transformation of anatase TiO2 to rutile TiO2 (Reaction
4) initiates very slowly and exothermally at around 910 K. Due to
the facts that, more TiO2(a) is being produced by both reactions
1 and 3 and it is an exothermal reaction in nature, the production
of rutile TiO2 occurs at a very fast rate in the last section of the kiln.
4.3. Parametric study
In this section, the effect of various parameters on the predicted
length of the kiln, rutile TiO2 production rate, axial solid and
gas temperature distributions, and axial variation in solids
274 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

Fig. 12. Axial composition profiles of various compounds.

Table 3a
Predicted Kiln Length and Rutile TiO2 Production Rate vs. Moisture Percent (Dry Basis).

Moisture Percent (Dry Basis) L1 (m) L2 (m) L3 (m) L = L1 + L2 + L3 (m) Rutile TiO2 Production Rate (kg/hr)
60.0 17.0 18.5 9.75 45.25 3649.68
75.14 17.25 21.0 7.5 45.75 3648.24
90.0 17.75 23.0 6.0 46.75 3646.80

Table 3b
Predicted Kiln Length and Rutile TiO2 Production Rate vs. Dry Solids Mass Flow Rate.

Dry Solids Mass Flow Rate (kg/s) L1 (m) L2 (m) L3 (m) L = L1 + L2 + L3 (m) Rutile TiO2 Production Rate (kg/hr)
1.0 14.25 19.75 9.75 43.75 2789.28
1.31 17.25 21.0 7.5 45.75 3648.24
1.6 20.0 21.75 6.25 48.0 4458.96

Table 3c
Predicted Kiln Length and Rutile TiO2 Production Rate vs. Gas Mass Flow Rate.

Gas Mass Flow Rate (kg/s) L1 (m) L2 (m) L3 (m) L = L1 + L2 + L3 (m) Rutile TiO2 Production Rate (kg/hr)
4.0 17.0 15.75 3.25 36.0 3641.76
6.218 17.25 21.0 7.5 45.75 3648.24
8.0 17.0 23.75 12.25 53.0 3651.12

Table 3d
Predicted Kiln Length and Rutile TiO2 Productiion Rate vs. Kiln Inclination Angle.

Inclination Angle (deg) L1 (m) L2 (m) L3 (m) L = L1 + L2 + L3 (m) Rutile TiO2 Production Rate (kg/hr)
1.4 16.25 20.0 7.75 44.0 3651.12
2.29 17.25 21.0 7.5 45.75 3648.24
5.0 18.75 23.25 7.5 49.5 3642.48
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
Table 3e
Predicted Kiln Length and Rutile TiO2 Production Rate vs. Kiln Rotational Speed.
Rotational Speed (r.p.m) L1 (m) L2 (m)
0.21 16.5 20.5
0.33 17.25 21.0
1.0 18.25 22.25
composition is discussed. In Tables 3a–3e, predicted kiln length
values and rutile TiO2 production rates for various parameters
are shown and the base values (Table 2) are highlighted in bold
letters.
4.3.1. Predicted kiln length as a function of various parameters
Table 3a reveals that, as the moisture percentage is increased
from 60 to 90, the lengths of the first and second sections of the
kiln increase. This is because of the fact that more amount of mois-
ture will require large amount of heat to reach the saturated state
and then to get vapourized. Therefore, lengths of the both sections
will be higher for this case. This rise in length of the second section
also causes delay in completion of Reaction 1 and 2 as they speed
up only after evaporation is finished. On the contrary, the length of
the third section decreases on increasing water content. This can
be explained as follows. Though it takes longer for water to get
vaporized, the fill-angle of the kiln and amount of dry solid mass
available in third section remains same and it has to meet the same
requirement of phase transformation in chemical reactions. Now,
total heat transfer to the solids in last section for the solid with
higher water content will be greater than that for solid with lower
water content. This is because of exposure of the dry solid to the
hotter gas in the former case as heating starts at a longer distance
from the solid inlet. The combined effect on total length is that it is
greater for higher moisture content case. Table 3b reveals that as
dry solids mass flow rate increases from 1.0 kg/s to 1.6 kg/s, pre-
dicted kiln length increases from 43.75 m to 48.0 m. This is
because higher solids residence time is needed to achieve the
desire objective and hence more kiln length is needed. Here, mois-
ture percentage remains same but amount of moisture present in
wet solids increases which raises the length of the second section.
However, the length of the third section decreases due to higher
heat transfer from gas and wall to solids. The overall effect is that
the predicted kiln length increases. As gas flow rate at inlet
increases from 4 kg/s to 8 kg/s, total kiln length increases from
36.0 m to 53.0 m (Table 3c). Near solids inlet, in spite of lower
gas residence time at higher gas flow rate, there is enough heat
transfer to the material to reach the boiling point of water in
almost same distance as in lower gas flow rate cases. However,
at latter stages of second section, effects of higher gas flow rate
are evident on processing of solid charge, as lower gas temperature
at the same axial location results in less heat transfer to solids, and
lengths of second and third section increase significantly. Table 3d
reveals that as the kiln inclination angle changes from 1.4° to 5°,
predicted kiln length increases from 44 m to 49.5 m. This is
expected because, as inclination angle increases, axial velocity of
solid also increases as V z ¼ U tan / a nd hence the residence time
of the solid is less. Hence, to reach the same composition of solids
at the exit, greater kiln length is needed. At higher rotational speed
of the kiln, the axial velocity of the solid increases and therefore,
the solid particles are exposed to gas for less amount of time. This
results in slower progression of reactions and therefore, to obtain
same solids chemical state of the product, length of the kiln has
to be increased (Table 3e).
4.3.2. Rutile TiO2 production rate as a function of various parameters
The rate of production of rutile TiO2 is calculated by dividing the
mass of rutile TiO2 produced, by the residence time of the solids in
275
L3 (m) L = L1 + L2 + L3 (m) Rutile TiO2 Production Rate (kg/hr)
8.25 45.25 3651.12
7.5 45.75 3648.24
7.25 47.75 3639.24
the kiln. Table 3a shows that with the increase in moisture percent
from 60 to 90 the rate of production of rutile Titanium dioxide
decreases slightly from 3649.68 to 3646.8 kg/hr. This is because
of higher kiln length required for drying in the case of larger water
content in the solids. On the other hand, as dry solids mass flow
rate increases from 1 kg/s to 1.6 kg/s the production rate substan-
tially rises from 2789.28 kg/hr to 4458.96 kg/hr (Table 3b). This is
because of higher amount of solid charge being processed and fas-
ter rate of final reaction per unit time. Table 3c indicates higher gas
flow rate results in larger production of rutile TiO2 per unit time.
This is because solid temperature is slightly higher in the third seg-
ment of the kiln which accelerates the rate of reaction. Increasing
kiln inclination angle and kiln rotational speed reduces the rutile
titanium dioxide production rate (Tables 3d and 3e) since the kiln
has to be longer and hence the solids residence time in the kiln is
more.
4.3.3. Axial solid and gas temperature distribution as a function of
various parameters
Fig. 13 shows that, with the increase in moisture content in wet
solids, axial gas temperature and @T s
increase in third section of the
@z
kiln while axial solid temperature with respect to percentage kiln
length decreases. The explanation of this trend is as follows. For
a larger amount of water in the solid charge, the wet solid will have
to travel a longer distance to be totally moisture-free. This implies
that the same amount of dry mass of solid will have to be heated to
the same exit temperature. It may be noted that the fill angle
remains the same in the third section no matter what the original
moisture content is. So, the dry solid will be exposed to hotter gas
in the case of high moisture content as the heating starts close to
the kiln exit which is the inlet for the hot gas. Therefore, to reach
the requisite exit temperature, the dry solid will have to travel a
smaller percentage length of the kiln and hence the axial temper-
ature gradient of the solid is larger. The axial solid temperature at
the same percentage kiln length will be obviously lower. Fig. 14
depicts that, for higher solid mass flow rate, gas temperature is
considerably higher in the third section of the kiln. This is because
for larger solid mass flow rate, the fill angle increases resulting in
reduction in the mean beam length as the gas volume decreases.
Hence, the loss of heat by the gas by radiation is smaller and hence
the gas temperature is high. Fig. 15 indicates that increasing gas
flow rate causes significant drop in gas inlet temperature. The rea-
son is that higher gas flow rate means less gas residence time and
hence the kiln has to be longer (see Table 3c). Thus, solid temper-
ature is higher at the same percent kiln length in the third segment
of the kiln. Gas on the other hand loses more heat to the solid and
hence the gas temperature drops in the last section of the kiln for
higher gas flow rates. Finally, lower kiln inclination angle and
lower rotational speed result in lower gas temperature and higher
solid temperature at the same percent kiln length in the third sec-
tion of the kiln (graphs not shown). This is because of more resi-
dence time of the solids in the kiln which leads to shorter kiln
length.
Finally, lower kiln inclination angle and lower rotational speed
result in lower gas temperature and higher solid temperature at
the same percent kiln length in the third section of the kiln (graphs
not shown). This is because of more residence time of the solids in
the kiln which leads to shorter kiln length.
276 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

Fig. 13. Axial solid and gas temperature distributions for different moisture percentages (dry basis).

Fig. 14. Axial solid and gas temperature distributions for different mass flow rates of the dry solids.

4.3.4. Axial composition of solid charge as a function of various
parameters
Fig. 16 shows the effect of moisture content on the percent con-
version of Anatase TiO2. It is revealed that both Reactions 1 and 2
(Table 1a) progress faster as solids enter the third and the last
section of the kiln. As the third section comes latter for higher
moisture content case, these reactions along with the
decomposition of TiO.SO4 to TiO2 (a) and final reaction of
conversion occur later and relatively faster due to higher gas
temperature.
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279 277

Fig. 15. Axial solid and gas temperature distributions for different mass flow rates of the gas.

Fig. 16. Axial profiles of percent conversion of anatase TiO2 to rutile TiO2 for different moisture percentages (dry basis).

Fig. 17 depicts how solids mass flow affects the percent conver- and second sections because of larger solid mass flow rate and also
sion of anatase TiO2. At higher dry solids mass flow rates Reaction 3 larger amount of moisture to be evaporated per unit time. How-
and the final reaction are delayed due to requirement of longer first ever, at the same percent kiln length near the exit reactions occur
278 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279

Fig. 17. Axial profiles of percent conversion of anatase TiO2 to rutile TiO2 for different mass flow rates of the solids in the last 20% of the kiln length.

Fig. 18. Axial profiles of percent conversion of anatase TiO2 to rutile TiO2 for different mass flow rates of the gas in the last 20% kiln length.
 A. Agrawal, P.S. Ghoshdastidar / International Journal of Heat and Mass Transfer 106 (2017) 263–279
at a faster rate as compared to the cases of lower dry solids flow
rates.
Fig. 18 shows the influence of gas flow rate on the percent con-
version of anatase TiO2. The conversion of anatase TiO2 to rutile
TiO2 starts earlier in the kiln for lower gas flow rate due to higher
inlet gas temperature resulting in higher solids temperature.
Figs. 16–18 also depict that the final reaction in the kiln, or, con-
version of anatase TiO2 to rutile TiO2 takes place in approximately
last 12% of the kiln length.
For higher kiln inclination angles and kiln rotational speeds the
reaction process of conversion from anatase TiO2 to rutile TiO2
starts earlier with respect to percent kiln length (graphs not
shown) because kiln length is larger and hence the solids temper-
ature is higher near the solids outlet. The reaction rate is also faster
for higher kiln inclination angles and kiln rotational speeds. It may
be noted that the reaction rate should not be confused with pro-
duction rate of rutile TiO2 discussed in Section 4.3.2.
5. Conclusions
This paper presents a computer simulation study of calcination
of hydrous titanium dioxide for the production of rutile titanium
dioxide in rotary kiln. The heat transfer model includes turbulent
gas convection, radiation heat exchange among the hot gas, refrac-
tory wall and the solid surface, conduction in the refractory wall,
and mass and energy balances in the gas and solids. The gas con-
vection has been simulated using a stand-alone CFD model assum-
ing the kiln to be non-rotating and solid bed to be stationary. Based
on the aforesaid CFD data the local convection heat transfer coeffi-
cients have been calculated. The present computer model predicts
steady-state axial distributions of solids and gas temperature that
are in excellent agreement with an earlier work [10]. The axial
chemical composition profile of the solid charge is consistent with
the trends found in literature. The kiln length predicted by the pre-
sent model is 45.75 m as compared to 45 m of the actual kiln
reported in Ginsberg and Modigell [10]. The accurate prediction
of the kiln length is a major achievement of this work. The capabil-
ity of the present model to predict the kiln length and the mod-
elling of gas convection are two novel aspects of this work which
can greatly help in the design and optimization of such kilns. A
detailed parametric study reveals that higher moisture content in
the solids results in slightly larger kiln length and lower rutile-
TiO2 production rate. However, higher dry solids mass flow rate
requires slightly larger kiln length but gives rise to substantial
rutile-TiO2 production rate. For low speed gas flow kiln length is
much smaller although the rate of production of rutile titanium
dioxide is marginally lower. Smaller kiln inclination angles and kiln
rotational speeds result in shorter kiln length and higher rutile-
TiO2 production rate. On the whole the parametric study lent a
good physical insight into the rutile titanium dioxide production
process in a rotary kiln.
References
[1] A. Saas, Simulation of the heat transfer phenomenon in a rotary kiln, Ind. Eng.
Chem. Proc. Des. Dev. 6 (4) (1967) 532–535.
[2] F.A. Kamke, J.B. Wilson, Computer simulation of a rotary dryer-II. Heat and
mass transfer, AIChE J. 32 (2) (1986) 269–275.
[3] C.A. Cook, V.A. Cundy, Heat transfer between a rotating and a moist granular
bed, Int. J. Heat Mass Transfer 38 (3) (1995) 419–432.
[4] P.S. Ghoshdastidar, V.K. Anandan Unni, Heat transfer in the non-reacting zone
of a cement rotary kiln, J. Eng. Ind., Trans. ASME 118 (1996) 169–172.
279
[5] P.S. Ghoshdastidar, G. Bhargava, R.P. Chhabra, Computer simulation of heat
transfer during drying and preheating of wet iron ore in a rotary kiln, Drying
Technol. 20 (1) (2002) 19–35.
[6] P.S. Ghoshdastidar, A. Agarwal, Simulation and optimization of drying of wood
chips with superheated steam in a rotary kiln, J. Therm. Sci. Eng. Appl., Trans.
ASME 1 (2009) 024501.
[7] X.Y. Liu, J. Zhang, E. Specht, Y.C. Shi, F. Herz, Analytical solution for the axial
solid transport in rotary kilns, Chem. Eng. Sci. 64 (2009) 428–431.
[8] R. Schmidt, P.A. Nikrityuk, Numerical simulation of the transient temperature
distribution inside moving particles, Can. J. Chem. Eng. 90 (2012) 246–262.
[9] Y. Sonavane, E. Specht, Numerical analysis of the heat transfer in the wall of
rotary kiln using finite element method ANSYS, in: Proceedings of 7th
International Conference on CFD in the Minerals and Process Industries,
CSIRO, Melbourne, Australia, 2009.
[10] T. Ginsberg, M. Modigell, Dynamic modelling of a rotary kiln for calcination of
titanium dioxide white pigment, Comput. Chem. Eng. 35 (2011) 2437–2446.
[11] G. Dumont, P.R. Bélanger, Steady state study of a titanium dioxide rotary kiln,
Ind. Eng. Chem. Proc. Des. Dev. 17 (2) (1978) 107–114.
[12] J. Roubal, D. Pachner, V. Havlena, T. Hanis, Analysis of the rotary kiln model, in:
Proceedings of European Control Conference, Budapest, Hungary, 2009.
[13] P. Koukkari, I. Laukkanen, K. Penttila, New applications with time-dependent
thermochemical simulation, in: Proceedings of 3rd Colloquium on Process
Simulation, Espoo, Finland, 1996.
[14] H.A. Spang, A dynamic model of a cement kiln, Automatica 8 (1972) 309–323.
[15] A. Manitius, E. Kurcyusz, W. Kawecki, Mathematical model of the aluminium
oxide rotary kiln, Ind. Eng. Chem. Proc. Des. Dev. 13 (2) (1974) 132–142.
[16] H.K. Guruz, N. Bac, Mathematical modelling of rotary cement kilns by the zone
method, Can. J. Chem. Eng. 59 (1981) 540–548.
[17] A.P. Watkinson, J.K. Brimacombe, Limestone calcination in a rotary kiln, Metall.
Trans. B 13 (1982) 369–378.
[18] A.P. Watkinson, J.K. Brimacombe, Oxygen enrichment in rotary lime kilns, Can.
J. Chem. Eng. 61 (1983) 842–849.
[19] J. Mu, R.A. Hard, Simulation study of a phosphate nodulizing kiln, Ind. Eng.
Chem. Proc. Des. Dev. 23 (1984) 374–381.
[20] P.S. Ghoshdastidar, C.A. Rhodes, D.I. Orloff, Heat transfer in a rotary kiln during
incineration of solid waste, in: Proceedings of 23rd ASME National Heat
Transfer Conference, Denver, Colorado, 1985, 85-HT-86.
[21] R.T. Bui, G. Simard, A. Charette, Y. Kocaefe, J. Perron, Mathematical modeling of
the rotary coke calcining kiln, Can. J. Chem. Eng. 73 (1995) 534–545.
[22] R.A. Davis, Mathematical model of magnetite oxidation in a rotary kiln furnace,
Can. J. Chem. Eng. 74 (1996) 1004–1009.
[23] R.A. Davis, D.J. Englund, Model and simulation of a ported kiln for iron oxide
pellet induration, Can. J. Chem. Eng. 81 (2003) 86–93.
[24] M. Georgallis, P. Nowak, M. Salcudean, I.S. Gartshore, Modelling the rotary
lime kiln, Can. J. Chem. Eng. 83 (2005) 212–223.
[25] F. Marias, H. Roustan, A. Pichat, Modelling of a rotary kiln for the pyrolysis of
aluminium waste, Chem. Eng. Sci. 60 (2005) 4609–4622.
[26] F. Mintus, S. Hamel, W. Krumm, Wet process rotary cement kilns: modeling
and simulation, Clean. Technol. Environ. Policy 8 (2006) 112–122.
[27] K.S. Mujumdar, V.V. Ranade, Simulation of rotary cement kilns using a one-
dimensional model, Chem. Eng. Res. Des. 84 (A3) (2006) 165–177.
[28] K.S. Mujumdar, K.V. Ganesh, S.B. Kulkarni, V.V. Ranade, Rotary Cement Kiln
Simulator (RoCKS): integrated modeling of pre-heater, calciner, kiln and
clinker cooler, Chem. Eng. Sci. 62 (2007) 2590–2607.
[29] D.K. Fidaros, C.A. Baxevanou, C.D. Dritselis, N.S. Vlachos, Numerical modelling
of flow and transport processes in a calciner for cement production, Powder
Technol. 171 (2007) 81–95.
[30] K. Shahriari, S. Tarasiewicz, Modelling of a clinker rotary kiln using operating
functions concept, Can. J. Chem. Eng. 89 (2011) 345–359.
[31] H. Ratajska, A. Przepiera, M. Wisniewski, Thermal transformations of hydrous
titanium dioxide, J. Therm. Anal. 36 (1990) 2131–2134.
[32] W.F. Sullivan, S.S. Cole, Thermal chemistry of colloidal titanium dioxide, J. Am.
Ceram. Soc. 42 (3) (1959) 127–133.
[33] H.C. Hottel, Radiation heat transfer, in: W.H. McAdams (Ed.), Heat
Transmission, 3rd ed., McGraw-Hill, New York, USA, 1954, Chap. 4.
[34] J.K. Brimacombe, A.P. Watkinson, Heat transfer in a direct-fire rotary kiln-I.
Pilot plant and experimentation, Metall. Trans. B 9 (1978) 201–208.
[35] V. Gnielinski, Neue Gleichungen für den Wärme- und Stoffübergang in
turbulent durchströmten Rohren und Kanälen, Forsch. Ingenieurwes. 41 (1)
(1975) 8–16.
[36] S.H. Tscheng, A.P. Watkinson, Convective heat transfer in a rotary kiln, Can. J.
Chem. Eng. 57 (4) (1979) 433–443.
[37] J.K. Brimacombe, A.P. Watkinson, Heat transfer in a direct-fired rotary kiln-II.
Heat flow results and their interpretation, Metall. Trans. B 9 (1978) 209–219.
[38] N.V. Suryanarayana, J.E. Lyon, N.K. Kim, Heat shield for high temperature kiln,
Ind. Eng. Chem. Proc. Des. Dev. 25 (1986) 843–849.
[39] H. Helmrich, K. Schugërl, Rotary kiln reactors in the chemical industry, Ger.
Chem. Eng. 3 (1980) 194–202.
[40] M.W. Chase, NIST-JANAF thermochemical tables, J. Phys. Chem. Ref. Data
Monogr. 9 (1998) 1–1951.