You are on page 1of 14

ARTICLE IN PRESS

Chemical Engineering Science 65 (2010) 28212834

Contents lists available at ScienceDirect

Chemical Engineering Science


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

A coupled uid dynamic-discrete element simulation of heat and mass


transfer in a lime shaft kiln
T. Bluhm-Drenhaus , E. Simsek, S. Wirtz, V. Scherer
Department of Energy Plant Technology, Ruhr-University of Bochum, Universitaetsstrasse 150, D-44780 Bochum, Germany

a r t i c l e in fo

abstract

Article history:
Received 1 February 2009
Received in revised form
10 January 2010
Accepted 15 January 2010
Available online 21 January 2010

In the present study heat and mass transfer related to the chemical conversion of limestone to
quicklime in a shaft kiln are investigated by means of a coupled numerical scheme for gas and solid
phase transport. The three-dimensional transport of mass, momentum and energy in the gas phase is
modelled by computational uid dynamics (CFD), while a discrete element method (DEM) is employed
for the mechanical movement and the conversion reactions of the solid material. The DEM simulation
readily describes the mechanical and thermal particle-to-particle interactions of a large number of
differently sized particles. Novel aspects addressed in this work are the simultaneous effects of inner
particle heat-conduction and pore-diffusion of the gaseous product of the calcination reaction (CO2)
modelled by a shrinking core approach. Simulations of laboratory scale experiments of single reacting
spheres show good agreement with the measured conversion rates. Simulations of an idealised vertical
shaft kiln including pressure drop calculations demonstrate the suitability of the proposed approach for
the modelling of industrial scale systems.
& 2010 Elsevier Ltd. All rights reserved.

Keywords:
Shaft kiln
Discrete element method
Heat transfer
Granular materials
Porous media
Powder technology

 Mass transfer of the gaseous CO2 from the reaction front to the

1. Introduction
In lime industry natural limestone is taken from a quarry and
granulated. Subsequently it is calcined in an endothermic process
within a lime kiln. The main component of limestone calcite
(calcium carbonate) is dissociated at high temperatures according
to the following reaction (Stark and Wicht, 2000):
CaCO3 s-CaOs CO2 g

DHR 178 kJ=mole

A reaction front advances towards the centre of the CaCO3


particle and an outer shell of highly porous CaO is formed, see
Fig. 1. In the present work this effect is tackled by a shrinking core
approach with separate energy balances for the core and the shell
of each particle.
The dissociation temperature depends on factors like origin
and composition of the limestone, crystal lattice, grain size and
partial pressure of CO2 in the surrounding atmosphere (Boynton,
1980).
Factors limiting the reaction rate can be classied into three
categories (Oates, 1998):

 Heat transfer from the surrounding gas phase to the reaction




front including convection to the surface and conduction


through the oxide shell,
Chemical kinetics of the dissociation reaction,

 Corresponding author. Tel.: + 49 234 32 26333; fax: + 49 234 32 14227.

E-mail address: bluhm-drenhaus@leat.ruhr-uni-bochum.de


(T. Bluhm-Drenhaus).
0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.01.015

surrounding including diffusion through the product shell and


the boundary layer of the gas phase.
Recent work by other authors concerning this eld of research
mainly has dealt with one-dimensional approaches for the
simulation of vertical shaft kilns. Shagapov and Burkin (2008)
analysed chemical and physical kinetics in a coke red mixed feed
vertical shaft kiln by mathematical modelling. Bes (2006)
developed a numerical model for a vertical shaft kiln, which
takes into account details of heat and mass transfer using
different gaseous and solid fuels for the calcination. Other work
has concentrated on global energy balances and generalised
process simulations, which rely on eld tests to provide the
necessary input data. Thermodynamic analysis of an annular shaft
kiln was performed by Senegacnik et al. (2008) investigating the
effect of excess air ratio and ue gas recirculation. Twodimensional coupled DEM-CFD simulations of heat transfer in a
rotary kiln with inert spheres have been presented by Shi et al.
(2008). To the authors knowledge no coupled DEM-CFD simulations of lime shaft kilns have been published in the available
literature.

2. Particle-ow coupling
To illustrate the various interactions of a limestone particle
with its surroundings, Fig. 2 outlines the processes taking place on
an exemplary particle track through a shaft kiln.

ARTICLE IN PRESS
2822

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

CO2

mGas,out

TGas,out

H2O
CaCO3

Tp

CaO

Fig. 1. Shrinking core model of calcium carbonate dissociation.

In the upper part of the kiln the particle is heated by


convection and the water content evaporates. In the central part
energy is supplied by an external source, e.g. combusting fuel and
the calcination process releases carbon dioxide. In the lower part,
the particle is cooled by the counter-owing air coming from the
bottom of the kiln and the air is consequently preheated. A
pressure drop is induced by the ow through the void fraction of
the packed bed. Hence a detailed model of the movement and
reaction of the granular phase as well as a model for the ow of
the gas phase is needed to simulate a lime shaft kiln. Finally the
interactions between granular matter and uid ow have to be
described.
The reacting granular phase is modelled with a DEM code,
which is described in Section 3. It has been developed at the
Department of Energy Plant Technology, University of Bochum, in
earlier work (Kruggel-Emden et al., 2007a; Simsek et al., 2008a, b).
The simulation of the uid phase is conducted with the
commercial CFD code FLUENT, which applies a nite-volume
method for the solution of conservation equations for mass,
momentum and energy of a continuous uid (Chorin, 1968;
Patankar, 1980). A pressure based solver with the SIMPLE
pressure velocity coupling is used (Fluent, 2006).
In this paper a quasi steady-state simulation is discussed. It
aims at the stationary mode of operation of a shaft kiln with
constant inlet and outlet temperatures and gas concentrations.
The simulation is conducted in two stages. In the rst stage
particle motion due to the force of gravity is computed
independently of the uid ow, heat transfer and reactions. In
the second stage these effects are included in a coupled
simulation of particle conversion and uid ow on the basis of
the particle tracks resulting from stage one. This separation is
applicable because of two characteristic features of the lime
conversion process. Namely there is no signicant acceleration of
the lime particles in the moving bed due to uid ow and there is
no signicant change in diameter due to chemical reaction. The
calcination reaction gives rise to a mass transfer from the solid to
the uid phase, but the simultaneously increasing particle
porosity justies the assumption of constant diameters (Stark
and Wicht, 2000). Due to these facts the particle motion is
independent of uid ow and so the two stages can be computed
sequentially.
The advantage of this two-staged scheme is a considerable
saving of computational resources, since the mechanical simulation of the particle motion consumes most of the simulation time
due to small time steps usually required for DEM simulations
(Kruggel-Emden, 2007). Note that the rst stage (the calculation
of the particle motion) of a lime kiln simulation containing
approximately 18,000 spherical particles (see Section 5), lasts
about one week on a multi-processor parallel computing cluster.

CO2

Qexternal

Tp

mGas,in

TGas,in

Fig. 2. Example of a particle track in a shaft kiln with heat, mass and momentum
transfer.

The second stage, the coupled simulation of particle reaction and


uid ow, requires just about one day of computational time.
Therefore, parameter variations for different modes of thermal
operation of the shaft kiln (e.g. different heat input, mass ow or
temperature of gas phase) can be quickly simulated based on the
set of pre-computed particle tracks.
Although the coupled simulation of the solid phase and the
uid ow is less time consuming than the computation of the
particle tracks, the details of the coupling are of great importance
for a reliable and converging solution.
In the DEM and the CFD simulation different grid types are
employed. The computational background grid for the DEM
simulation is of orthogonal, Cartesian type. This grid is mainly
used to narrow down the search-space for contact detection
between particles. On the contrary, the unstructured, nonorthogonal CFD mesh denes the spatial discretization of the
underlying differential equations (Fluent, 2006).
Coupling between the solid (particles) and uid (gas) phase is
carried out using a particle source in cell (PSIC) approach.
Volume-averaged sources of mass, energy and momentum
resulting from the drying, heating and calcination processes are
transferred to the mesh of the gas phase simulation for each cell.
Feedback from the uid to the solid phase is provided by local
boundary conditions of interpolated gas temperature, ow
velocity and gas composition on the solids side. This two-way
information exchange is communicated via a transfer grid, which

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

DEM

Transfer Grid

2823

CFD

particle
sources

fluid
properties

Fig. 3. Interphase feedback via a transfer grid.

is used for interpolation and data transfer between the two


simulation programs, see Fig. 3.
The transfer grid is necessary because of the size range of the
particles in relation to the resolution of both the DEM and the CFD
grids. The CFD cells are of the same order of magnitude as the
particles diameters, while the DEM cells are much larger.
The transfer grid has a substantially smaller spatial resolution
than both the DEM grid and the CFD mesh, to capture the effect of
particles overlapping different uid cells. Furthermore,
the identication of the particle positions in the Cartesian
transfer grid is much more efcient than in conventional CFDmeshes, which might contain tetrahedral or other geometric
elements.
For the transfer of mass and energy to or from the gas phase
the sources of a single particle are distributed over the transfer
cells occupied by this particle, see top of Fig. 3. Consequently they
are applied to the associated uid cells, which share a common
region with the transfer cells. A transfer cell is associated to
exactly one CFD cell. For the assembly of particles in the packed
bed all sources are summed up for the respective transfer cells. If
CFD cells partially overlap different particles, each cell will receive
a fraction of the particle sources according to the number of
associated transfer cells.
In the opposite direction the uid properties are associated to
the respective transfer cells and here they are employed as local
boundary conditions for the particle simulation, see bottom of
Fig. 3. In this simple interpolation scheme the conditions of a
transfer cell are applied to a single particle, if the particles centre
of gravity lies within this transfer cell. This implies the assumption that local variations in the uid properties are negligible
compared to the particles diameter.
Table 1 shows general phenomena which occur in coupled
simulations of uid ow through packed beds and related
quantities, which determine the feedback between solid and
uid phase and vice versa.
For pressure drop calculations an isotropic porous medium is
assumed for the packed bed. The macroscopic ow is assumed to
be laminar, because the resolution of the CFD mesh does not
capture any effect of turbulence in the cavities between the

Table 1
Parameters inuencing the coupled DEM-CFD simulation.
Solid phase (DEM)

Fluid phase (CFD)

Phenomenon

Particle heating
Evaporation of water
Chemical conversion

Gas cooling
Flow distribution
Pressure drop

Interphase feedback

Energy sink
Mass source (CO2, H2O)
Solid/void fraction

Gas temperature
Gas composition
Relative velocity

particles. The inuence of turbulence is modelled on a sub-grid


scale in the context of the pressure drop and heat transfer
submodels. Note that this is a common approach to handle uid
ow in packed beds of low permeability (Fluent, 2006).
Coupling parameters which are relevant to convergence
control are the number of iterations performed by the uid
and the solid branch of the simulation between each interval of
data transfer and the under-relaxation of the source terms,
respectively.

3. Submodels
The submodels described in the following sections have been
chosen to account for a sufcient representation of the underlying
phenomena, while being simple enough to limit the computational effort for systems with large numbers of particles.
A major simplication is the assumption that the limestone
particles are of spherical shape. This simplication is justied to a
certain extend by the fact that the surface of lime stones particles
is smoothed and rounded during their travelling time within a
shaft kiln (typically several hours) due to mechanical interaction
of the particles with each other and hence approach a spherical
shape. The consideration of shape details would have led to an
unacceptable numerical effort.
Note that the pressure drop correlations used here are adapted
to non-spherical particles by a shape factor. The inuence of

ARTICLE IN PRESS
2824

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

particle shape on inner particle heat conduction and diffusion is


not included at the current state and is subject of further research.
3.1. Particle motion
The description of mechanical motion is based on the standard
DEM approach as introduced by Cundall and Strack (1979) solving
the Newtons and Eulers equations:
mi

N
X
d2~
xi
~

g;
F ij mi ~
2
dt
j1

Ji

N
X
~i
d2 j
~ ij

M
2
dt
j1

with mass m, moments of inertia Ji , accelerations d2~


x i =dt 2 , angular
~
~ i =dt 2 , external forces ~
F and external moments M
accelerations d2 j
describing the translational and rotational motion. When all
forces are determined, the equations of motion can be integrated
numerically. For this purpose several integration schemes are
applicable (Rougier et al., 2004). Because of its simplicity and low
computational effort, the single step explicit forward Euler
scheme with constant time steps has been chosen, which turned
out to be sufcient for many engineering applications (KruggelEmden, 2007).
The normal force during contact of two particles or a particle
and a wall can be modelled by a simple approach based on the
n
v . The following force law is
overlap d and the normal velocity ~
applied in the current work for normal forces (Simsek et al.,
2009):
n
n
n
n
~
F el ~
F diss kn d~
n gn ~
v
F ~
n

3
n

k is the stiffness of a linear spring and g is a damping coefcient.


For tangential forces the force law holds:
t
n
~
F j~
x =j~
xj
F minkt x; mC j~

k is the stiffness of a linear spring, mC is the Coulomb friction


x is the relative tangential displacement. The
coefcient and ~
tangential force is the minimum of frictional force and the force of
the tangential elastic spring. Although being a simple force model,
a combination of a linear spring and Coulomb friction allows for a
good representation of macroscopic collision properties like
tangential coefcients of restitution and rebound angles.
A review and evaluation of normal force models can be found
in Kruggel-Emden et al. (2007b) and for tangential force models in
Kruggel-Emden et al. (2007c).
3.2. Interphase heat transfer
As stated in the introduction, the heat transfer from the uid to
the solid phase is dominated on the one hand by convection from
the uid to the particle surface and on the other hand by
conduction within the calcium oxide shell of the lime particles.
Radiative transfer and particleparticle conduction are relatively
small in packed beds of low thermal conductivity, because
neighbouring particles have small temperature differences especially, if the particle mixing rate is low as it is the case in shaft
kilns. Therefore, radiation and particleparticle conduction have
been neglected in the current work. However, these effects have
been implemented within a DEM code for grate ring systems by
our group (Simsek et al., 2009).
The convective heat transfer is based on established Nusseltcorrelations for single spherical particles with a laminar and a
turbulent contribution (Baehr and Stephan, 2006). A correction
factor for packed beds of granular media according to Gnielinski
(1980) is applied:
Nubed fe Nusphere

fe 1 1; 51e

In Eq. (6) e is the void fraction of the packed bed and it is valid in
the range 0.26o e o1. The result is a mean Nusselt number for a
region of constant void fraction, which yields the mean heat
transfer coefcient:

am

Nubed  lg
dp

In principle the heat conduction within the particle should


account for two effects. Firstly, there is a reaction front travelling
through the particle creating a porous outer shell of CaO with low
thermal conductivity. Secondly, the introduction of a heat sink by
the endothermic calcination reaction of CaCO3 is required. This
typically leads to a hot shell of CaO during heat up and a rather
cold core of CaCO3. Due to the typical size of lime stone particles
in the order of several centimetres a temperature prole within
the particle is expected during heat up and cooling (also without
any calcination reaction), because Biot numbers (Eq. (8)) are
typically above one.
Bi

a  dp
lp

Rth;i
Rth;a

Only for Biot numbers much less than unity, which holds true for
small particle diameters or high thermal conductivities, the heat
transfer is dominated by the outer thermal resistance due to
convection (1/a) and a uniform temperature throughout the
particle can be assumed (Baehr and Stephan, 2006). For porous
lime particles typically present in shaft kilns with an effective
relative gas velocity of about 9 m/s, the Biot number is larger than
unity, as can be seen in Table 2.
The Biot number effect could be taken into account by
calculating a temperature distribution as a function of the radius.
This would require a detailed discretization of the particle which
is very time consuming, when large numbers of particles are of
interest as in technical systems. A simplied approach for the
radial temperature distribution, which can be applied in DEM
simulations based on an analytical approach, has been presented
by Rickelt et al. (2009). However, because for real scale lime shaft
kilns the calculation of several hundred thousands of particles is
nally required, we decided to neglect the Biot number effect in a
rst step and concentrate on modelling of the travelling reaction
front within the particle as the dominating effect for the heat
balance within the particle. A shrinking core approach is applied
in the current work.
As there are two distinct shells, calcium carbonate and calcium
oxide, which each possess constant material properties, we
assume that a combined energy balance for each shell (S1, S2) is
suitable for the description of the thermal behaviour.
The energy balance for the innermost shell (core, S1) holds:
mS1 cS1

dTS1
kS12 TS2 TS1 DHR
dt

The change of the inner energy of the core results from conductive
transfer from the outer shell and the reaction enthalpy of the
endothermic calcination reaction. A threshold value is applied for
the minimum core mass to eliminate the singularity when the
particle is completely converted to calcium oxide. The heat

Table 2
Biot number of typical lime particles.

a
lp,CaO
dp
Bi

W m2 K1
W m1 K1
m

227.5
0.7
0.01
3.25

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

transfer coefcient (kS12) from shell 1 to shell 2 for a spherical


particle is determined by a series of the two different effective
thermal conductivities.
kS12

lS1

1
rS1

4plS1 lS2



1
1
 rS2;m
 r1S1
lS2 rS1;m

during the important stage of conversion. The enthalpy of reaction


(178 kJ/mol) dominates the overall energy balance, since it is
more than twice the amount of energy required for heating the
limestone from 20 to 900 1C.

10
3.3. Reactions and interphase mass transfer

The difference between the mean radii of each shell, see Fig. 4, is a
measure for the mean path length of conductive transfer.
The energy balance for the outer shell (S2) is accordingly:
mS2 cS2

2825

dTS2
am Ap TG TS2 kS12 TS2 TS1
dt

11

Based on Eqs. (9) and (11) the spatial variation of temperature


inside the particle, which would in reality be a continuous prole
over the radius, is reduced to two average temperatures for the
inner (S1) and outer (S2) region. The change of inner energy of the
outer shell (Eq. (11)) results from convective transfer from the gas
phase and conductive transfer to the core. Note that a low
conductive heat transfer coefcient (kS12) in relation to the
convective transfer coefcient am (i.e. Bi 41) leads to an
accumulation of energy in the shell, so that the shell temperature
is close to the gas temperature.
At the beginning of the calculation a particle consists of pure
calcium carbonate and a uniform temperature exists throughout
the particle. Once the reaction starts (the particle gets heated
along its pathway through the lime shaft kiln) a porous calcium
oxide shell is formed and heat is conducted further into the core
of the particle, where the energy is consumed by the reaction.
Thus the temperature of the calcium carbonate core is lower than
the temperature of the outer calcium oxide shell. Note that we
have assumed the average value of the two shell temperatures as
the temperature at the reaction front, which is essential to the
calculation of reaction progress, as shown in the next section.
In the late stages of conversion the heat transfer through the
shell can be a limiting factor for the reaction. When the particle is
fully decomposed, the core vanishes and only the energy balance
for the shell applies with a spatial uniform temperature for the
remaining calcium oxide particle. Hence, the shrinking core
approach only accounts for inner particle temperature gradients

rS2

The thermal decomposition of calcium carbonate is controlled


by heat transfer, reaction kinetics or mass transfer phenomena,
depending on the specic boundary conditions of the process. The
slowest transfer mechanism determines the overall reaction rate.
As a result, the calcination temperature stabilizes at a specic
value where a local equilibrium exists between energy input from
the uid phase and energy consumption by the endothermic
reaction.
Heat transfer dominates the conversion in case of large
particles ( 10  2 m), high reaction temperatures and low partial
pressure of CO2 in the gas phase, whereas at low temperatures
and high CO2 partial pressures mass transfer through the
product layer and through the particle boundary layer control
the reaction rate. In small particles ( 10  5 m), as present in
pulverised limestone, the chemical resistance becomes important
(Stark and Wicht, 2000). Particle sizes in typical shaft kilns lie in
the range of several centimetres. Due to this fact, the present work
concentrates on the rst two mechanisms, while the latter is
neglected in this context.
A shrinking core model (Mohr, 2001) is adopted for the
calcination process. The reaction rate is proportional to the area of
the reaction front, the difference between the CO2 concentration
at the reaction front cCO2 ;s1 and outside the uid dynamic
boundary layer cCO2 ;1 and a series of resistances Ri,C:
@mCaCO3
cCO ;s1 cCO2 ;1
p  d2CaCO3  MCaCO3  P22
@t
i 1 Ri;C;ST

In the regime of large particles and intermediate partial pressure


of CO2 the reaction is inhibited by two mass transfer resistances
(R1,C, R2,C) for the diffusion of CO2 to the gas phase. At the same
time the thermal resistance of the CaO shell limits the heat
transfer to the reaction front as stated above.
Equilibrium partial pressure (Eq. 13) is assumed for the
concentration at the reaction front. It is strongly dependent on
the local temperature at the reaction front, which in turn results
from the energy balance, see Section 3.2.
cCO2 ;S1

rS2m

CaO
rS1
rS1m

12

pCO2 ;eq
R  Ts1

13

The CO2 equilibrium pressure is calculated by the following


expression, see Fig. 5.


19 140
Pa
14
pCO2 ;eq 1:2  1012 exp 
Tg
The reaction starts, when a minimum temperature is reached,
where the equilibrium concentration at the interface exceeds the
concentration of CO2 in the surrounding of the particle.
The rst diffusion resistance originates from the porous
calcium oxide layer (S2):

CaCO3
R1;C

1 ds1  ds2 ds1



2 ds2  Deff

15

CO2 ;CaO

Fig. 4. Shrinking core model for heat transfer with mean radii indicated.

When the reaction proceeds, the thickness of the product layer


(ds2  ds1) increases and so does the diffusion resistance. Here the
effective diffusion coefcient is dominated by the Knudsen
diffusivity (Cussler, 1984; Veldsink et al., 1995):


eCaO
1
1 1
Deff
2

16
i;CaO
Dmol;i DKn;i
t

ARTICLE IN PRESS
2826

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

1.00E+06

CO2 equilibrium pressure [Pa]

1.00E+05

1.00E+04

1.00E+03

1.00E+02

1.00E+01

1.00E+00
200

300

400

500
600
700
Temperature [C]

800

900

1000

Fig. 5. CO2 equilibrium pressure at the reaction interface.

For the temperature dependent molecular diffusion coefcient of


CO2 in air a polynomial function is applied. The Knudsen diffusion
coefcient is dened as
s
8RTp
2
DKn;i r pore
17
pMi
3

Erguns equation has been modied for polydisperse beds of


non-spherical particles by introducing the volume weighted
diameter for differently sized particles:
"
#1
N 
X
Vi 1

dp
21
V dpi
i1

The Knudsen diffusivity is inuenced by the mean pore radius,


which is a function of the specic inner surface of calcium oxide
SCaO.

The shape factor FD accounts for the effect of the irregular


curvature of the particles surface. It can be assessed experimentally from plots of pressure drop versus supercial uid velocity
for a given granular material. In general, packed beds which have
the same void fraction and mean diameter, but possess a different
packing lattice (e.g. cubic, cell centred or irregular) or consist of
different materials also have different pressure loss characteristics (Molerus, 1993), so the shape factor has to be quantied for
each case.
The advantage of Erguns equation is its mathematical
simplicity which leads to a more stable behaviour of the
numerical solution. In our scheme, the pressure drop of the uid
phase is calculated independently from the particle motion,
because the packed bed is dense and features an approximately
uniform distribution of the void fraction. The particle velocity is
very slow (  1 m/h) compared to the gas velocity (  1 m/s)
leading to a relative velocity that can be assumed to be equal to
the gas velocity as in a xed bed.

r pore;CaO 2

eCaO
SCaO  r0CaO

18

The second diffusion resistance describes the effect of the


boundary layer diffusion from the particle to the gas phase. It
depends on the Sherwood number Sh (Baehr and Stephan, 2006).
R2;C

d2s1
Sh  dp  DCO2 ;Gas

19

The strongest impact on the overall reaction in this context results


from the limiting effect of the Knudsen diffusion through the
porous lime layer. The key material parameter is the specic inner
surface, which is different for lime from different origins and
changes due to sintering effects. In our model we use a constant
value at this stage, but the effect of sintering is subject of further
research.

4. Results on laboratory scale calcium carbonate


decomposition

3.4. Interphase momentum transfer


In the present work the pressure drop across the packed bed is
calculated by a modied Ergun equation (Ergun, 1952). The uid
motion through the packed bed is not deterministic and an
isotropic porous medium is assumed instead. Basically, the
correlation for pressure difference per unit length of the
percolated packed bed is a square function of the supercial
u gas ~
u particle j with some semi-empirirelative uid velocity U0 j~
cal coefcients derived for the packed bed. These depend on the
particle diameter and the void fraction as characteristic measures
for the particle bed and on the dynamic viscosity and the density
of the uid.
jDpj
150m  1e2
1:75r  1e 2
U0
U0

L
FD  d p 2  e3
FD  d p  e3

20

Hills (1968) performed thermo-gravimetric analysis of the


decomposition of single sintered spheres of calcium carbonate. He
proposed a reaction mechanism similar to the one utilised in our
work and he gave experimental evidence for the hypothesis that
the process was controlled by heat and mass transfer for large
particles. Simulations of the laboratory scale apparatus shall be
compared with the experimental results as a verication of the
proposed model.
4.1. Experimental setup
The weight loss of spherical carbonate samples was measured
in a high temperature furnace tube by a micro balance, which was
mounted on a water cooled plate above the furnace. Air and

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

air/CO2 mixtures were swept through the furnace at a constant


temperature. Simultaneously, the core temperature of the particle
was measured by a thin wire thermocouple, which was placed
inside the sample and the wires were attached to the balance. The

samples with an outer diameter of approximately 10 mm were


prepared by compacting and sintering analytic-reagent grade
calcite powder.
Each run was started with the sample at thermal equilibrium,
as pure CO2 was swept through the furnace during heat-up of the
sample and furnace (pure CO2 suppresses the reaction at
temperatures below 900 1C). Then the CO2 was abruptly changed
to the desired air/CO2 mixture to initiate the decomposition.

Table 3
Particle properties.
dp
SCaO

mm
m2 kg  1

W m1 K1

W m1 K1

t
eCaO
lCaO

eCaCO3
lCaCO3

2827

107 0.2
16 000
1.41
0.56
0.7
0.04
2.26

4.2. Decomposition rates


The experimental results were evaluated by plotting the
dimensionless radius r* of the reaction front position versus time.
When all calcium carbonate is consumed, the dimensionless

Table 4
Gas phase conditions for the experimental runs (Hills, 1968).
Run
Temperature (1C)
CO2 mole fraction (-)

26
877
0

28
876
0.106

30
874
0.154

35
877
0.323

41
902
0

Run 35 meas.

Run 30 meas.

Run 28 meas.

Run 26 meas.

Run 35 sim.

Run 30 sim.

Run 28 sim.

Run 26 sim.

45
829
0

52
720.5
0

1
0.9
0.8
0.7
r* [-]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20

30

40
50
Time [min.]

60

70

80

Run 35 meas.

Run 30 meas.

Run 28 meas.

Run 26 meas.

Run 35 sim.

Run 30 sim.

Run 28 sim.

Run 26 sim.

90

1
0.9
0.8
0.7
c* [-]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20

30

40
50
Time [min.]

60

70

80

90

Fig. 6. Runs conducted at constant temperature and different CO2 concentrations (see Table 4), A: dimensionless radius (r*), B: calcination degree (c*), measured data
shown as symbols, simulation data as lines.

ARTICLE IN PRESS
2828

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

radius approaches zero.




r
mtm1 1=3
r
m 1=3
m0 m1
r0

The relation of r* can also be transformed to a dimensionless


degree of conversion (calcination degree c*):
22
c 1r  3

These plots would show straight lines, if the reaction was only
controlled by a chemical step proportional to the surface area of
the reaction front. Mass conservation considerations lead to the
molar reaction rate:
n_ const  4pr 2 

dr
dt

25

As the diameter of the sample is given in dimensionless form only


in the referenced publication and the true diameter of 10 mm is
given without a range of uncertainty the sensitivity of the model
on changes of particle diameter has been assessed. In conclusion
we derived diameters within a range of 1070.2 mm which tted
the experimental results of each run. The particle properties
presented in Table 3 were used for the simulation.
The value of SCaO was assumed by Hills for the evaluation of his
experiments and was consequently adopted for the simulations.
Table 4 gives an overview of the set of parameters varied in the
experiments.
The results of the simulations in comparison with the
experimental data are shown in the following gures.
Fig. 6 shows that at a constant temperature of approximately
877 1C the concentration of CO2 in the gas phase inuences the
duration of the calcination process, because the concentration
gradient across the product shell and the boundary layer as a

23

For constant dr/dt at constant temperature, this equation shows


that the reaction rate per unit area of the reaction interface would
also be a constant. So an Arrhenius-type reaction rate could be
assumed.
dr
n_

dt
const  4pr 2

m0 mt
m0 m1

24

Since the plots of r* versus time exhibit a slight s-shape in the


experiment, especially in the late stages of the decomposition, the
Arrhenius assumption is not true in this case and Hills, in contrast,
derives the diffusion/heat transfer model for the process, which
has been adopted in our work.
Run 52 meas.

Run 45 meas.

Run 41 meas.

Run 26 meas.

Run 52 sim.

Run 45 sim.

Run 41 sim.

Run 26 sim.

1
0.9
0.8
0.7
r* [-]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20

30

40
50
Time [min.]

60

70

80

Run 52 meas.

Run 45 meas.

Run 41 meas.

Run 26 meas.

Run 52 sim.

Run 45 sim.

Run 41 sim.

Run 26 sim.

90

1
0.9
0.8
0.7
c* [-]

0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20

30

40
50
Time [min.]

60

70

80

90

Fig. 7. Runs conducted at different temperatures in pure air (see Table 4), A: dimensionless radius (r*), B: calcination degree (c*), measured data shown as symbols,
simulation data as lines.

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

2829

880
Ts1 sim.

Ts2 sim.

T core meas.

Temperature [C]

860

840

820

800

780

760
0

10

20

30

40
Time [min.]

50

60

70

80

Fig. 8. Simulated core (Ts1) and shell (Ts2) temperature during the decomposition of a 10 mm sample in a furnace held at 863 1C, measurement data (core) as symbols.

driving force decreases with decreasing CO2 gas phase


concentration.
Fig. 7 highlights that in pure air the temperature level
determines the rate of calcination. At higher temperatures the
equilibrium pressure of CO2 at the reaction front (Eq. (13)) rises
and the concentration gradient towards the surrounding gas
phase increases.
The core temperature of the sample during calcination (Fig. 8)
shows a characteristic development. It starts at gas phase
temperature, because the experiment begins in thermal
equilibrium. Then the core temperature sharply decreases due
to the fast onset of the endothermic reaction. It reaches a
minimum when the energy consumed by the reaction equals
the energy transferred to the reaction front by conduction. In the
later stages the oxide shell increases in thickness and the reaction
is limited by the mass transfer of CO2 from the particle to the
surrounding gas phase. The end of the reaction is marked by an
unsteady jump of the core temperature to the constant gas
temperature, as soon as the calcium carbonate is completely
converted to calcium oxide and the core radius approaches zero.
This specic development of the core temperature provides
experimental evidence for the proposed shrinking core model.
The good agreement between calculations and measurements
conrms that the current model approach is appropriate for the
simulation of larger CaCO3 particles. This conrmation was the
basis for further numerical studies in a semi technical shaft kiln.

5. Simulation of a lime shaft kiln


A scaled down vertical shaft kiln is selected for numerical
investigation of the heat and mass transfer phenomena, because it
is the simplest conguration comparable to technical kilns, which
include a preheating, a calcinating and a cooling zone. Using this
simplied example the performance of the three dimensional
simulation method is studied. In consequence, the method can be
employed to other (full scale) kiln types, which feature more
complex ow patterns of gas or solid and thus are not accessible
by simple one dimensional approaches.

Assumptions made for the simulation of the vertical shaft kiln


are:

 The feed stone contains 98.5 wt% calcite (CaCO3) and 1.5 wt%
moisture.

 The gas is treated as an ideal mixture containing N2, O2, CO2


and H2O.

 The heat transfer between gas and solid in the packed bed is
dominated by convection.

 The heat supplied by combustion is treated in a simplied way





as a volumetric source of energy which does not release


additional CO2.
Spherical particles are assumed for contact detection between
particles.
A shape factor accounts for the inuence of irregular particles
on pressure drop (FD = 0.6, derived from measurements).

Note that in the current paper a simplied approach for the


supply of thermal energy to the kiln has been selected. Firstly, the
total amount of air supplied to the kiln is not split into cooling air
and secondary combustion air, so all of the air travels through the
cooling zone. Secondly, the thermal power is added by a
volumetric heat source and not by the combustion of a fuel as it
takes place in real shaft kilns. We selected this approach because
it was our main intention to test the derived calcination model
in a polydisperse, multi-particle, semi-technical reactor. The
consideration of the additional production of CO2 and consumption of O2 introduced by a combustion process could either be
taken into account by calculating the equilibrium production of
CO2 connected to the generation of unit amount of heat or by
introducing a gas phase combustion model. In this case, the
number of degrees of freedom in the boundary conditions would
be higher and would introduce additional uncertainties, which are
not desired at the current state of the development. However, no
principle obstacles exist to include a combustion model into the
simulations and this inclusion is subject of current research.
The impact of the simplied heat source and hence the lack of
CO2 from combustion is rather small for the overall process. The
effect of a smaller CO2 concentration in the surrounding of the

ARTICLE IN PRESS
2830

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

Table 5
Particle properties.
dp
SCaO

t
eCaO
lCaO

eCaCO3
lCaCO3

mm
m2 kg  1

W m1 K1

W m1 K1

17y45
5000
1.41
0.56
0.7
0.04
2.26

particles (cCO2 ;1 in Eq. (12)) results in a shift of the reaction


temperature to lower values and an over prediction of the
reaction rate, because the reaction rate is directly proportional
to the concentration difference between the particle and the uid.
As can be seen from Fig. 5 the equilibrium pressure exceeds one
atmosphere above 900 1C, so that there is a positive concentration
difference for any possible CO2 concentration in the gas phase.
Material properties of the limestone and quicklime are given
in Table 5.
The particle size distribution is represented by four discrete
classes of 17, 22, 32 and 45 mm. A specic surface of about
5000 m2/kg was found by Moropoulou et al. (2001) for a particular
limestone by mercury intrusion porosimetry and adsorption of
nitrogen according to the BET method. The effective thermal
conductivity of CaO lies in a range from 0.5 to 1.0 W m  1 K  1
(Cheng et al., 2007; Oates, 1998). All particle properties still bear a
high degree of uncertainty as there is substantial variation for
different origins of limestone.

5.1. Boundary conditions


By the counter-current ow of gas and solid in a vertical shaft
kiln internal heat recovery is achieved to some extent, since the
ue gases from the calcination zone preheat the feed limestone
and the calcined quicklime preheats the combustion air. This heat
recovery is limited by two major imbalances in the heat transfer
requirements (Oates, 1998):

(a) The kiln gases contain considerably more heat than is


required to preheat the limestone to calcination temperature
(approximately 900 1C).
(b) The quicklime does not contain sufcient heat to preheat the
air required for combustion (including excess air) to 900 1C.

This imbalance can be reduced by transferring heat form the


exhaust gases to the combustion air by external heat recovery.
The simplied geometry is shown in Fig. 9, see also Fig. 3. It is
scaled down by a factor of 1:5 compared to a common shaft kiln
with 10 m of shell height.
The scaled kiln is operating with the basic parameters
summarised in Table 6.
A rst reference simulation was carried out using cooling air at
an inlet temperature of 20 1C. The thermal input of 80 kW was set
to meet the theoretical energy requirement for the heating, drying
and calcination processes. The heat is supplied at the horizontal
mid-plane (height 1 m) of the kiln (the 80 kW are equivalent to a
ring rate of about 13.1 kg/h of lignite).
The second simulation was subsequently run with increased
inlet air temperature (350 1C), so that the thermal input (also
80 kW) was supplied at a higher temperature level. This measure
would be realized in actual plants by installing an air pre-heater
which transfers energy from the tail end ue gases to the inlet air.

Fig. 9. Geometry of the simulated vertical shaft kiln with input streams.

Table 6
Operating conditions.

Height
Diameter
Angle of cone
Input (100% CaCO3)
Residence time
Air ow rate
Stone inlet temperature
Air inlet temperature
Thermal input

m
m
1
t d1
h
m3/h STP
1C
1C
kW

Run 1 (base)

Run 2 (air preheat)

2
0.64
60
3.29
8
233.5
20
20
80

2
0.64
60
3.29
8
233.5
20
350
80

Table 7
Results for the gas phase.

Mean gas velocity at inlet


Mean gas velocity in the
calcination zone (height 1 m)
Pressure drop

Run 1
(base)

Run 2
(preheated air)

m s1
m s1

4.02
0.73

8.55
0.88

Pa

2135

4064

5.2. Results
The main characteristics of the resulting gas ow are
summarised in Table 7 for both runs.
The mean supercial gas velocity magnitude at the air inlet is
much lower in run 1 than in run 2, due to the different densities at
the respective inlet temperatures. The velocity in the lower part of
the calcination zone is quite similar in both cases, because the
temperature levels are almost the same in this position. The
pressure drop resulting from the overall ow pattern in run 2 is
almost doubled compared to run 1. It is mainly determined by the
high velocity at the tip of the cone (air inlet).
Fig. 10 shows the particle residence times in steady state
operation. The mean residence time of approximately 8 h
(28,800 s) is met by most of the particles. Some at the outer
wall have longer and some in the core of the granular ow have
shorter travelling times through the kiln.

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

2831

the calcination zone, because additional energy is supplied by air


pre-heating.
The shell temperatures, see Fig. 12, in the upper part of the kiln
are equal to the core temperatures, since the calcium oxide
product layer is initially not present or very thin. The water

Fig. 10. Particle residence time (A: run 1, B: run 2).

Fig. 12. Particle shell temperature (Ts2, A: run 1, B: run 2).

Fig. 11. Particle core temperature (Ts1, A: run 1, B: run 2).

The temperatures of the particle cores resulting from the


applied model for shell conduction are shown in Fig. 11. It is
obvious that the particles approach thermal equilibrium shortly
after passing the calcination zone. While travelling towards the
particle outlet they have adopted the temperature of the counter
owing gas phase. In run 2 higher temperatures are obtained in

Fig. 13. Particle calcium oxide mass fraction (A: run 1, B: run 2).

ARTICLE IN PRESS
2832

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

content evaporates, while the particles fall into the kiln. In the
calcination zone the shell temperatures are substantially higher
than the core temperatures of the particles. The temperature at
the reaction interface develops from energy consumption by the
calcination reaction and energy supply by convection through the
boundary layer to the shell surface and conduction through the
outer shell. In the lower part of the kiln the particles are again in
thermal equilibrium with the cooling air.
The result of greatest interest from the current simulation is
the degree of calcination of the particles, which is proportional to
the calcium oxide mass fraction. These results are presented in
Fig. 13.
It can be seen that the kiln is operated at very poor efciency in
run 1. The temperature level in the burning zone is generally too
low to achieve complete conversion, see Fig. 14. Nevertheless the
ue gases still contain a high amount of energy. The overall mass
based degree of calcination (c*) is only 55% in run 1.

Fig. 14 shows the composition of the quicklime particles at the


outlet of the kiln separately for each class of particle diameter, for
this purpose a sample of 10,000 particles was selected in steady
state operation. Small particles of 17.0 mm contain a high amount
of CaO (97.6 wt%), while larger particles show a successively
smaller degree of conversion (i.e. only 26.6 wt% CaO for 45.0 mm
particles). This correlation shows the effect of thermal resistance
in the outer shell, which inhibits conduction of heat towards the
centre of the particle.
As the energy of the ue gases in run 2 is used to pre-heat the
inlet air, a higher temperature level and a higher degree of
conversion are achieved, see Fig. 15. The overall degree of
calcination (c*) in run 2 is about 90%.
As can be seen in Fig. 15 the small particles (diameter 17.0 and
22.4 mm) are completely calcined (100 wt% CaO). The larger
particles (diameter 31.5 and 45.0 mm) contain signicant
amounts of unconverted CaCO3 (5.6 and 29.3 wt%, respectively).

100%
CaO
CaCO3

90%

particle composition [wt.-%]

80%
70%
60%
50%
40%
30%
20%
10%
0%
17.0

22.4
31.5
particle diameter [mm]

45.0

Fig. 14. Particle outlet composition by size (run 1, sample N = 10,000).

100%
CaO

90%

CaCO3

particle composition [wt.-%]

80%
70%
60%
50%
40%
30%
20%
10%
0%
17.0

22.4
31.5
particle diameter [mm]

Fig. 15. Particle outlet composition by size (run 2, sample N = 10,000).

45.0

ARTICLE IN PRESS
T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

2833

This behaviour can be explained when visualizing the


temperature proles along the centreline of the kiln, see Fig. 16.
The temperatures where rapid conversion occurs lie above 850 1C
(indicated in Fig. 16). As can be seen, in run 2 a larger number of
the particles exceed this limit than in run 1.
Due to air pre-heating a higher temperature in the calcination
zone is achieved in run 2. However, the higher maximum
temperature may have a negative inuence on the product
quality, when producing soft burned lime. The outlet gas
temperature is very similar in both cases.
The heat and mass balance of the kiln including input and
output streams is summarised in Table 8.
The energy imbalance of about 1% results from unsteady
variations in the particle ow, which slightly disturbs the steady
state of the gas phase simulation.
A crucial point of kiln operation is maintaining a high degree of
conversion while not over-burning the lime with excessive heat
or too long residence times at high temperatures. The product
quality decreases in this case due to sintering, which diminishes
the specic surface and reactivity of the quicklime. This occurs
rst with smaller particles while at the same time large particles
are not yet thoroughly calcined. Manufacturers of lime kilns have
developed different concepts to meet the requirements for soft or
hard burned lime. So a compromise has to be found in kiln design
and operation to meet an optimum regarding energy efciency
and product quality. As shown, detailed numerical simulations
may help to understand the underlying physical phenomena and
to optimise kilns during the design stage or to propose well
directed adaption of the operation parameters in response to
changing boundary conditions or material properties.

6. Conclusion

Fig. 16. Proles of particle and gas temperature along the centreline of the kiln,
(A: run 1, B: run 2).

Heat and mass transfer in a lime shaft kiln are investigated


by means of a coupled numerical scheme for uid and solid
phase transport. The strength of the coupled CFD-DEM simulation
lies in the detailed and particle centred description of momentum-, heat- and mass-transfer phenomena for a large number of
individual particles including particle-particle interaction.
The effects of interphase heat transfer and chemical conversion
(calcination) were studied in simulations of laboratory scale
experiments of single reacting spheres. The results were in good
agreement with the measured decomposition rates. The inuence
of inner particle heat-conduction and pore-diffusion of gaseous
product (CO2) was incorporated in a shrinking core model.

Table 8
Heat and mass balance (reference temperature 25 1C).
Run 1
Mass (kg h
INPUT
Feedstone, wet
Gas
Total
OUTPUT
Quicklime, dry
Gas
Energy source
Evaporation
Reaction
Total

Run 2
1

Mass (kg h  1)

Temperature (1C)

Energy (kJ s  1)

 0.2
 0.4
 0.6

138.96
301.21
440.17

20.0
350.0

 0.2
28.9
28.7

88.97
351.03

350.0
476.6

440.0

6.2
 48.6
80.0
 1.3
 51.4
 27.5

0.2
0.1%

1.2
1.5%

Temperature (1C)

Energy (kJ s

138.96
301.21
440.17

20.0
20.0

105.19
334.78

20.0
451.0

439.97

 0.1
 43.4
80.0
 1.3
 34.0
1.4

Imbalance

0.2
0.1%

0.8
1.0%

Calcination

55%

1

90%

ARTICLE IN PRESS
2834

T. Bluhm-Drenhaus et al. / Chemical Engineering Science 65 (2010) 28212834

Computation of particle motion and pressure drop in the


packed bed was additionally considered for the simulation of a
generic, semi-industrial vertical shaft kiln. The heat released
by combustion was modelled in a simplied manner as a
volumetric source of energy in the gas phase. In spite of the
applied simplications the results show a realistic behaviour of
kiln operation. Thus the procedure developed gives detailed
insight into physical phenomena related to kiln performance
and it opens up new possibilities to optimise different modes
of operation. Based on the results presented, extensions towards a
more realistic simulation including combustion are initiated.

t
FD

tortuosity, dimensionless
shape factor, dimensionless
angle, 1

Acknowledgement
This work was kindly supported by Thyssen-Krupp Technologies Polysius AG, Germany.
References

Notation
A
Bi
c
ci
c*
Di
d
F
f
g
HR
J
k
kS12
L
M
Mi
mi
_
m
N
Nu
n_
p
Q_

area, m2
Biot number, dimensionless
specic heat, J/(kg K)
molar concentration, mol/m3
calcination degree
diffusion coefcient, m2/s
diameter, m
force, N
factor, dimensionless
gravitational constant, m/s2
enthalpy of reaction, J/mole
momentum of inertia, kg m2
constant of stiffness
thermal conductivity, W/K
length, m
momentum, Nm
molar mass, kg/mol
mass, kg
mass ow, kg/s
number, dimensionless
Nusselt number, dimensionless
molar reaction rate, mol/s
pressure, N/m2
heat ow, J/s

R
Rth
Ri,C
S
Sh
r
r*
T
t
u, v, w
V
x, y, z
y

universal gas constant, J/(mol K)


thermal resistance, m2K/W
calcination resistance, s/m
specic surface area, m2/kg
Sherwood number, dimensionless
radius, m
dimensionless diameter, dimensionless
temperature, 1C or K
time, s
velocity, m/s
volume, m3
coordinate, m
normal vector, dimensionless

Greek letters

a
g
d
D

e
l

m
m
x

r
r0

heat transfer coefcient, W/(m2K)


damping coefcient
overlap, m
difference
void fraction, porosity, dimensionless
heat conductivity, W/(mK)
friction coefcient
dynamic viscosity, kg/(m s)
relative tangential displacement
density, kg/m3
bulk density, kg/m3

Baehr, H.D., Stephan, K., 2006. Warme-und Stoffubertragung,


fth ed. Springer, Berlin.
Bes, A., 2006. Dynamic Process Simulation of Limestone Calcination in Normal
Shaft Kiln, Ph.D. Thesis, Magdeburg.
Boynton, R.S., 1980. Chemistry and Technology of Lime & Limestone. Wiley,
New York.
Cheng, C., Specht, E., Kehse, G., 2007. Einuss der Herkunft und Stoffwerte von

Kalksteinen auf ihr Zersetzungsverhalten in Schachtofen.


ZKG International
60, 5161.
Chorin, A.J., 1968. Numerical solution of NavierStokes equations. Mathematics of
Computation 22, 745762.
Cundall, P.A., Strack, O.D.L., 1979. A discrete numerical model for granular
assemblies. Geotechnique 29, 4765.
Cussler, E.L., 1984. Diffusion: Mass Transfer in Fluid Systems. Cambridge
University Press, Cambridge.
Ergun, S., 1952. Fluid ow through packed columns. Chem. Eng. Prog. 48, 8994.
Fluent 2006. FLUENT 6.3 Users Guide, Fluent Inc., Lebanon NH.

Gnielinski, 1980. Warme-und


Stoffubertragung
in Festbetten. Chem. Eng. Technol.
52, 228236.
Hills, A.W.D., 1968. The mechanism of the thermal decomposition of calcium
carbonate. Chem. Eng. Sci. 23, 297320.
Kruggel-Emden, H., 2007. Analysis and Improvement of the Time-driven Discrete
Element Method, Ph.D. Thesis, Bochum 2007.
Kruggel-Emden, H., Simsek, E., Wirtz, S., Scherer, V., 2007a. A comparative
numerical study of particle mixing on different grate designs through the
discrete element method. J. Pressure Vess. Technol. 129, 593600.
Kruggel-Emden, H., Simsek, E., Rickelt, S., Wirtz, S., Scherer, V., 2007b. Review and
extension of normal force models for the discrete element method. Powder
Technol., 171173.
Kruggel-Emden, H., Simsek, E., Wirtz, S., Scherer, V., 2007c. A comparison and
validation of tangential force models for the use within discrete element
simulations, DEM07 Conference, Brisbane, 2729 August.
Mohr, M., 2001. Numerische Simulation der simultanen Reaktion von Kalkstein
und Kohle bei der Zementherstellung, Ph.D. Thesis, Bochum 2001.
Molerus, O., 1993. Principles of Flow in Disperse Systems. Chapman & Hall, London.
Moropoulou, A., Bakolas, A., Aggelakopoulou, E., 2001. The effects of limestone
characteristics and calcination temperature to the reactivity of the quicklime.
Cem. Concr. Res. 31, 633639.
Oates, J.A.H., 1998. Lime and Limestone. Wiley-VCH, Weinheim.
Patankar, S.V., 1980. Numerical Heat Transfer and Fluid Flow. Hemisphere,
Washington, DC.
Rickelt, S., Kruggel-Emden, H., Wirtz, S., Scherer, V., 2009. Simulation of heat
transfer in moving granular material by the discrete element method with
special emphasis on inner particle heat transfer. In: Proceedings of the 2009
ASME Summer Heat Transfer Conference, San Francisco, 1923 July.
Rougier, E., Munjiza, A., John, N.W.M., 2004. Numerical comparison of some
explicit time integration schemes used in DEM, FEM/DEM and molecular
dynamics. Int. J. Numer. Methods Eng. 61, 856879.
Senegacnik, A., Oman, J., Brane, S., 2008. Annular shaft kiln for lime burning with
kiln gas recirculation. Appl. Therm. Eng. 28, 785792.
Shagapov, V.Sh., Burkin, M.V., 2008. Theoretical modeling of simultaneous
processes of coke burning and limestone decomposition in a furnace. Combust.
Explos. Shock 44 (1), 5563.
Shi, D., Vargas, W.L., McCarthy, J.J., 2008. Heat transfer in rotary kilns with
interstitial gases. Chem. Eng. Sci. 63, 45064516.
Simsek, E., Brosch, B., Wirtz, S., Scherer, V., 2008a. Numerical Simulation of Solid
Fuel Conversion on a Forward Acting Grate with a Coupled CFD/Discrete
Element Method (DEM), 8th European Conference on Industrial Furnaces and
Boilers, Vilamoura, 2528. March.
Simsek, E., Grochowski, R., Wirtz, S., Scherer, V., Kruggel-Emden, H., Walzel, P., 2008b.
An experimental and numerical study of transversal dispersion of granular material
on a vibrating conveyor. Particulate Sci. Technol. 26, 177196.
Simsek, E., Brosch, B., Wirtz, S., Scherer, V., 2009. Numerical simulation of grate
ring systems using a coupled cfd/discrete element method (DEM). Powder
Technol. 193 (3), 266273.

Stark, J., Wicht, B., 2000. Zement und Kalk, Birkhauser,


Basel.
Veldsink, J.W., van Damme, R.M.J., Versteeg, G.F., van Swaaij, W.P.M., 1995. The use
of the dusty-gas model for the description of mass transport with chemical
reaction in porous media. Chem. Eng. J. 57, 115125.

You might also like