You are on page 1of 17

Int. J. Miner. Process.

83 (2007) 99 – 115
www.elsevier.com/locate/ijminpro

On the drying rates of individual iron oxide pellets


T. Tsukerman a , C. Duchesne a,⁎, D. Hodouin b
a
Département de Génie chimique, Université Laval, Québec, Québec, Canada G1K 7P4
b
Département de Génie des mines, de la métallurgie et des matériaux, Université Laval, Québec, Québec, Canada G1K 7P4
Received 26 June 2006; received in revised form 22 May 2007; accepted 22 June 2007
Available online 5 July 2007

Abstract

After agglomeration, iron oxide pellets are sintered in continuous furnaces to develop the mechanical properties required by iron
making plants. In the first zones of the furnace, pellets are dried by the hot recycled gas. The objective of the study is to model their
drying kinetics. For that purpose, individual pellets, instrumented for temperature measurement, are dried in a laboratory furnace
equipped with a thermo-balance. The results show that there are four stages in the drying process: 1—an evaporation of the water film
at the pellet surface; 2—an hybrid regime with surface film evaporation and apparition of dry spots with evaporation fronts moving
within the pellet; 3—a shrinking wet core leaving behind the evaporation front a dry shell where there is water vapour diffusion; 4—a
change of diffusion into bulk transportation of water through the dry shell, when the evaporation front is close to the boiling point.
Mass and heat transfer equations are numerically solved, and the simulated values compared to the experimental results.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Drying; Iron oxide pellets; Parameter estimation

1. Introduction ration is typically performed in moving grate or grate-


kiln furnaces, through which pellets are sequentially
Iron oxide pellets are one of the main feed sources for dried, fired, and cooled by direct contact with hot gases
blast furnaces or direct reduction processes of the iron of varying flow rates and temperatures. Pellets quality
and steel making industry. Pelletizing is the first step in (i.e. mechanical strength) strongly depends on their
the production of these pellets. It basically consists of residence time and temperature profile within the firing
mixing finely ground iron ore concentrate with several zone as well as their residual moisture content at the
additives and water in pelletizing disks or drums until time pellets reach the firing zone, since induration reac-
balls of a specific size are obtained (called green pellets). tions only begin when drying is completed. Pellet drying
Prior to shipment to steel plants, green pellets need to be is therefore a bottleneck for increasing production rate
fired in order to increase their mechanical strength for and pellet quality, and also is a major energy consumer
transportation and handling considerations. Pellet indu- (Thurlby and Batterham, 1980; Clark, 1981). It has been
reported that about 25% of the total amount of energy
required for pellet induration is used for drying (Patisson
⁎ Corresponding author. Tel.: +1 418 656 5184; fax: +1 418 656
et al., 1990). Thus, any small improvement in the drying
5993. performance of sintering furnaces can result in consid-
E-mail addresses: carl.duchesne@gch.ulaval.ca (C. Duchesne), erable savings for the pellet producer, both in capital and
daniel.hodouin@gmn.ulaval.ca (D. Hodouin). operating costs (Clark, 1981).
0301-7516/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.minpro.2007.06.004
100 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

Modelling of particular sections or complete indura- (1981), however, strongly suggest that these two regimes
tion furnaces has been a very active research area in are simultaneously active for most of the drying time, but
the 80's and 90's as a tool for improving the design of this observation has never been taken into account in any
such furnaces, optimizing their operating conditions and iron ore pellet drying models. Most published validation
minimizing energy consumption. The most complete results for single pellet drying only concentrate on fitting
models include heat and mass transfer between the gas weight loss versus time curves without taking into ac-
and the pellet bed, chemical reactions within pellets and count drying rate data and pellet temperature explicitly in
gas flow models in the various zones of moving grate model validation and/or parameter estimation (Young
machines (Voskamp and Brasz, 1975; Hasenack et al., et al. 1979; Thurlby and Batterham, 1980; Seshadri and
1975; Thurlby et al., 1979; Cumming et al., 1985; Cross da Silva Pereira, 1985). To the author's knowledge, the
and Wade, 1989; Cross and Blot, 1999; Pomerleau et al., work of Thurlby and Batterham (1980) is the only paper
2005), grate-kiln furnaces (Young et al., 1979; Thurlby, showing measurements of pellet temperature during
1988; Cross and Wade, 1989), and laboratory or pilot drying (taken at the pellet center). However, they have
scale pot-grate furnaces (Seshadri and da Silva Pereira, not used this information in the validation of their drying
1985; Cumming and Thurlby, 1990; Küçükada et al., rate expressions since heat transfer gradients within the
1992, 1994). Other modelling contributions were main- pellets were neglected. It will be shown in this paper that
ly focused on gas–solid heat transfer and chemical re- weight loss versus time curves are relatively insensitive
actions (Muchi and Higuchi, 1972; Dash et al., 1974; to the drying mechanism and that pellet temperature and
Davis, 1998) and drying (Thurlby and Batterham, 1980; drying rate data, in addition to weight loss data, are
Breiholtz and Hillberg, 1980; Ilic and Turner, 1986; necessary to discriminate drying mechanisms and to
Patisson et al., 1990) within pellet beds. improve the accuracy of drying rate expressions which,
In the past, several drying rate expressions have been in turn, should increase the accuracy of new furnace
developed but very few were validated on single pellets. designs or lead to improved operating conditions of
The simplest drying rate model assumes that evaporation existing ones when models are used for optimization.
occurs from the pellet surface in a single stage for the This paper therefore presents an experimental in-
entire drying time (Voskamp and Brasz, 1975). This vestigation of the drying of single iron oxide pellets
means capillary forces are such that internal water can followed by fundamental modelling of drying rates. It
rapidly and continuously supply the water film at the will be shown that using weight loss, drying rate, and
pellet surface to compensate evaporation until the pellet pellet temperature data all together in the development
is completely dry. However, the experimental investiga- of drying rate expressions leads to the conclusion that
tions of Clark (1981) and Nekrasova et al. (1988) on the both surface and internal drying occur simultaneously
drying mechanisms of iron pellets as well as the drying during a significant portion of total pellet drying time, as
rate curves published by Thurlby and Batterham (1980) was suggested by Clark (1981), and that a linear time
clearly show that drying of single iron ore pellets occur in varying combination of drying rates computed for each
several stages. Most drying models assume two stage regime provides a very good agreement against experi-
drying, involving surface evaporation until a critical mental drying data, and more accurate results than those
moisture content is reached, followed by internal drying previously published for single pellet drying.
(also called falling rate period) described using a shrink-
ing core modelling approach (Dash et al., 1974; Thurlby 2. Experimentals
et al., 1979; Young et al., 1979; Breiholtz and Hillberg,
1980; Thurlby and Batterham, 1980; Cumming et al., The drying behaviour of single iron ore pellets was
1985; Thurlby, 1988; Patisson et al. 1990; Küçükada investigated using thermo-gravimetry with forced flow
et al., 1992, 1994; Pomerleau et al., 2005). The existence of hot air. The pellets were prepared at the research
of a third drying stage (e.g. falling rate period is divided consortium COREM (Québec, Canada) according to a
in two stages) is also considered in a few papers (Muchi standard industrial recipe (major components: 65% total
and Higuchi, 1972; Hasenack et al., 1975; Seshadri and Fe, mainly Fe2O3 and small amounts of Fe3O4, 5%
da Silva Pereira, 1985) but is often neglected. SiO2, 8% water, 1.3 % C, 0.8% CaO–MgO), and were
A common assumption made in all modelling work balled to a mean diameter within the range of 30–
reviewed in this paper is that surface drying and internal 35 mm. Pellet size is therefore about three times larger
drying stages occur sequentially (i.e. a sharp transition than those produced in industry. However, this was
between the two regimes occur at the critical moisture necessary to allow the insertion of a few thermocouples
content). The experimental results published by Clark for studying temperature at different locations within the
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 101

pellets. It will be shown later that drying behaviour of within the pellets, the drag force exerted on thermo-
the studied pellets is very similar to that of industrial couples was introducing important variations (noise) in
pellets. the weight measurements due to the balance sensitivity.
The thermo-balance used in this study is shown in Therefore, weight loss and pellet temperature were
Fig. 1(a). Temperature within the chamber is automat- measured on different pellets dried under the same
ically controlled at the desired set-point using electrical conditions. A similar problem was encountered by
elements. Forced convection around the pellets is ob- Thurlby and Batterham (1980).
tained using a suction pump (see tube just above the The flow rate and temperature of the drying gas were
pellet in Fig. 1(a)). The volumetric air flow rate was selected to fall within the range of conditions typically
measured outside the chamber and also controlled at a encountered in industrial induration furnaces (Thurlby
desired value. The pellets are placed directly on the and Batterham, 1980; Clark, 1981). Moreover, it was
balance, which allow for continuous weight measure- decided to change gas flow rate and temperature in a
ment. Continuous recording of pellet temperature was correlated fashion in order to test increasingly aggres-
made using three thin thermocouples (0.5 mm), inserted sive drying conditions. For more details on the experi-
at the center, the middle, and at the surface of each ments performed in this work, the reader is referred to
pellet, as shown in Fig. 1(a) and (b). Pellet temperature Tsukerman (2006).
and weight data as well as gas flow rate and gas tem-
perature were collected every 10 s in each experiment. 3. Mathematical model for pellet drying
A total of six pellets were dried under three different
conditions as presented in Table 1. Pairs of pellets were A modified shrinking core modelling approach is
dried under the same conditions since it was impossible used to describe the drying behaviour of single pellets.
to record weight loss and pellet temperature simulta- The traditional shrinking core approach was used in
neously on each pellet with the thermo-balance used in nearly all drying models published for iron ore pellets. It
this work. Indeed, when the thermocouples are inserted basically assumes that drying occurs at the surface of the
pellet until a critical moisture Wpc content is reached,
below which the evaporation front starts moving within
the pellet (i.e. shrinking core) leaving a dry shell behind.
It is further assumed in the internal drying stage that
shrinking core moisture content is constant at Wpc until
drying is completed. The modification proposed in this
work consists in allowing both surface and internal
drying stages to occur simultaneously rather than se-
quentially (traditional approach). This section describes
the traditional shrinking core modelling approach for
iron ore pellets. Modifications to this model will be
discussed later, in the Results and discussion section.

3.1. Heat transfer within the pellets

Heat transfer within the dry shell is given by:


   
1 A 2 ATp ðr;t Þ dWp ðt Þ
kp;d 2 r þ
r Ar Ar dt
 
Vp ATp ðr; t Þ Aqp;d Cp;d Tp ðr;t Þ ð1Þ
 Cwv ¼
Ar Ar At

where λp,d, ρp,d, and Cp,d are the thermal conductivity,


the density, and the heat capacity of the dry shell, Cwv is
the heat capacity of water vapour, Tp, Wp, and Vp are the
pellet temperature, moisture content, and volume, and
Fig. 1. (a) Thermo-balance and (b) location of the three thermocouples finally, Ar is the surface area of a sphere of radius r. The
within the pellet. two terms on the left hand side of this equation
102 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

Table 1
Conditions of the drying experiments
Drying conditions set Monitored data Test Pellet diameter Pellet initial Gas temperature Gas flow rate
dp (mm) weight Mp,0 (g) Tg∞ (°C) G (l/min)
A Weight 1 30 54.5 60 40
Temperature 2 30 55.9 60 40
B Weight 3 32 55.4 100 60
Temperature 4 30 47.0 100 60
C Weight 5 30 55.3 150 80
Temperature 6 28 44.6 150 80

respectively consists of heat conduction within the dry water evaporation and to increase the sensible heat of the
shell (R b r b rc) and the sensible heat gained by the water wet pellet by heat conduction.
vapour, assumed to be at the same temperature as the  
pellet, when travelling from the evaporation front (r = rc)
to the pellet surface (r = R), and involves the drying rate
h Tgl ðt Þ  Tp ð R;t Þ  kp;w
ATp ðr;t Þ
Ar r¼R
j
expression (dWp(t) / dt) which will be discussed in the   1 dWp ðt Þ ð4Þ
 DH Tp ð R;t Þ ¼ 0:
next section. Ap dt
The energy conservation equation at the evaporation
However when only internal drying occurs (Wp ≤ Wpc),
front (r = rc) is given by Eq. (2), where the total amount
then all the heat transferred by the gas is used to increase
of heat supplied to the shrinking core by conduction
the sensible heat of the pellet via conduction through the
through the dry shell divides into conduction within the
dry shell (Eq. (5)).
wet core and latent energy used for water evaporation.
 
This boundary condition was also proposed by Cross
et al. (1979). kp;d
ATp ðr;t Þ
Ar j r¼R
¼ h Tgl ðt Þ  Tp ð R;t Þ : ð5Þ
 
kp;w
ATp ðr;t Þ
Ar j
rc 

þ DH Tp ðrc ;t Þ
 dWp ðt Þ Vp
dt Acore
The second boundary condition consists of symmetry
applied at the pellet center:

¼ kp;d
ATp ðr;t Þ
Ar j
rc þ
ð2Þ ATp ðr;t Þ
Ar j
r¼0
¼ 0: ð6Þ

Heat transfer within the wet core (rc b r b 0) is due to Finally, the initial conditions are
conduction only, as described by Eq. (3):
Tp ðr;0Þ ¼ Ta ð7Þ
 
1 A ATp ðr;t Þ Aqp;w Cp;w Tp ðr;t Þ where Ta is the ambient temperature.
kp;w r2 ¼ ð3Þ
r2 Ar Ar At
3.2. Mass transfer (drying rates)
Note that only Eq. (3) is required when surface drying
occurs alone (i.e. when Wp N Wpc) and a single wet When drying occurs at the surface (Wp N Wpc), only the
zone occupies the entire pellet, and when the pellets are convective mass transfer resistance is considered and the
completely dry (using physical properties of the dry drying rate expression is given by the following equation,
pellet λp,d, ρp,d, and Cp,d). which is the most commonly used expression to describe
Solving this system of PDEs (Eqs. (1)–(3)) requires the first stage drying (Dash et al., 1974; Voskamp and
two boundary conditions as well as an initial condition. Brasz, 1975; Hasenack et al., 1975; Thurlby et al., 1979;
For the surface boundary condition (r = R) two expres- Breiholtz and Hillberg, 1980; Seshadri and da Silva
sions are necessary depending on whether the pellet Pereira, 1985; Küçükada et al., 1992, 1994; Pomerleau
surface is wet or dry. When only surface drying occurs et al., 2005):
the film (or pellet surface) temperature Tp(R,t) should
dWp    
eventually reach the wet bulb temperature. The follow- ¼ akm Wge Tp ð R;t Þ  Wgl ð8Þ
ing heat balance (Eq. (4)) is therefore used to compute dt
the pellet surface temperature when Wp N Wpc, that is the where a is the pellet surface area per unit volume, km is the
heat transferred by convection from the gas is used for convective mass transfer coefficient, and W ge and W g∞
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 103

respectively correspond to the saturated (or equilibrium) when heat balance across the evaporation front (Eq. (2))
gas moisture concentration (calculated at the pellet surface leads to Tp(rc,t) ≥ 100 °C, then the drying rate becomes
temperature Tp(R,t)) and the moisture concentration in the driven by the hydrostatic pressure difference between the
bulk of the drying gas. evaporation front (equilibrium or saturation pressure Pe)
In the second drying stage, when the evaporation and the drying gas or atmospheric pressure (Patm). This
front moves within the pellet (Wp ≤ Wpc) and the evapo- pressure gradient generates a bulk flow of water vapour
ration temperature is below the normal boiling point that can generally be described by Darcy's law. Again
(Tp(rc,t) b 100 °C), it is normally assumed that drying is assuming that water vapour does not accumulate within
controlled by mass transfer (Hasenack et al., 1975; the dry shell (i.e. the water vapour exits the pellet at much
Seshadri and da Silva Pereira, 1985). In this work, it is faster rate than that of the receding front) and integrating
further assumed under those conditions that water va- the mass balance between the evaporation front and the
pour does not accumulate within the dry shell. In other pellet surface leads to the following expression (Tsuker-
words, the drying rate, the rate of diffusion of water man, 2006), used when Wp ≤ Wpc and Tp(rc,t) ≥ 100 °C:
vapour to the pellet surface, and the rate of convective
mass transfer at the pellet surface are all equal at any dWp 3KP ðPe  Patm Þ
¼  3  ð10Þ
dt R
time. The following drying rate expression is therefore
rc  R2
used where Dff is the effective diffusion coefficient of
water vapour within the dry shell of the iron ore pellet. where KP is the relative permeability coefficient of the
dry shell. Owing to the assumption that the wet core
dWp a Wge jr¼rc  Wgl
¼ 2 : ð9Þ remains at the critical moisture content Wpc until the
dt R Rrc þ 1 pellet is completely dry, the location of the evaporation
Rrc Dff km R 2
front can be calculated from a simple mass balance:
This expression, which simplifies to Eq. (8) when rc = R
 1=3
(i.e. surface drying), was also used in other studies Wp
(Seshadri and da Silva Pereira, 1985; Küçükada et al., rc ¼ R  : ð11Þ
Wpc
1992, 1994). This drying model is a simplification of
the transient one-dimensional Fick's law (ODE instead 3.3. Model parameters
of a PDE) used by Hasenack et al. (1975) to compute the
concentration profile of water vapour within the pellets. Most model parameters were obtained from tables or
This simplified approach may lump other phenomena published data (see the Appendix), because their values
(such as convective transport) when estimating Dff can be considered as reliable and well fitted to the studied
from drying data. This will be assessed and discussed material and to the applied experimental conditions.
later in Section 4. Other mathematical expressions for These are the heat capacities and densities of liquid water,
the falling rate period includes polynomial regression of silica, and hematite (other components were neglected)
drying curves (Patisson et al., 1990), multiplying the which are used to compute the dry and wet pellet densities
expression for the first stage drying by some function and heat capacities (ρp,d, ρp,w, Cp,d and Cp,w), the dry
involving current, critical and equilibrium moisture pellet thermal conductivity (λp,d) using the approach of
content (Dash et al., 1974), and assuming that a fixed or Akiyama et al. (1991, 1992), the heat capacity of water
a variable fraction of the total heat transferred to the vapour (Cwv), the latent heat of evaporation of water (ΔH)
pellet is used for evaporation (Muchi and Higuchi, and water vapour saturation pressure (Pe). The convective
1972; Thurlby et al., 1979; Young et al. 1979; Breiholtz heat and mass transfer coefficients (h and km) were calcu-
and Hillberg, 1980; Thurlby and Batterham, 1980). lated using standard correlations for spheres involving the
In industrial furnaces, the drying gas temperature is Reynolds, Prandtl, Schmidt dimensionless numbers. To
often higher than the normal boiling point of water. compute these numbers the wet and dry air thermal con-
When this temperature is reached at the evaporation front ductivity, viscosity and density were also necessary (see
(i.e. Tp(rc,t) = 100 °C), then it is usually assumed that the the Appendix).
drying rate is controlled by heat transfer (Hasenack et al., A total of six parameters were left for estimation
1975; Seshadri and da Silva Pereira, 1985). When heat using the experimental weight loss and pellet temper-
transfer rates are high, then evaporation rate at the front ature data. These parameters are listed below:
may exceed the rate at which the vapour leaves the pellet,
therefore resulting in an increased pressure within the • The critical humidity Wpc
pellet. To model such a situation, it was assumed that • The wet pellet thermal conductivity, λp,w
104 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

• A multiplicative empirical adjustment coefficient (α) analysis based on the normalized parameter sensitivity
for the correlation used to compute the convection matrix. In the iterative procedure proposed by Zhen
mass transfer coefficient of water vapour in the air et al. (2003), a parameter estimability order is estab-
film surrounding the pellet (see the Appendix) lished based on the ability of each parameter to predict
• The Reynolds number, which is used in the cor- the sensitivity matrix. This predictive ability is quanti-
relations for calculating the convective heat and mass fied by a residual sum of squares obtained through
transfer coefficients. It was decided to estimate this successive linear regression models (see Zhen et al.
parameter rather than calculating it, since fluid flow (2003) for more details). Table 2 shows the results for
around the pellets is complex (i.e. heterogeneous), the proposed drying model. In our case, Wpc is clearly
and gas flow is measured in a tube located outside the the most estimable parameter, followed by Re number.
drying chamber. Fluid flow heterogeneities around To determine the subset of estimable parameters, a
the pellets are therefore lumped into a single average cut-off value on the residual sum of squares is used.
Re number estimated from experimental data Zhen et al. (2003) suggest a general cut-off value of
• An adjustment factor β allowing the calculation of 0.04. However, when noise level is high, or when the
the effective water diffusivity in the pores of the dried model is imperfect, they rather suggest a cut-off value of
pellet part; this factor, which divides the water diffu- 0.3. In our case, a cut-off value of 0.3 leads to only two
sion coefficient to give the effective diffusivity, cor- estimable parameters (Wpc and Re), whereas five param-
responds to the ratio of the pore tortuosity to the pore eters are found to be estimable using a cut-off value of
volume fraction (see the Appendix) 0.04 (only the permeability coefficient KP is not
• The gas permeability of the dry pellet KP, required for estimable).
the Darcy's evaporation regime. This confirms the preliminary observations made
when trying to estimate the six parameters listed above.
3.4. Numerical solution of the system of equations It was found that the simulated results are not strongly
influenced by parameters λp,w and KP because, on one
The transient PDEs describing heat transfer within hand, the heat conductivity phenomena are fast com-
the dry and wet zones of the pellets were solved using a pared to the heat convection phenomena (low Biot
finite difference method using a time step of 0.1 s and a number), and on the other hand, for most experimental
radius step of 1 mm. These values were selected to tests, the pellet drying is terminated before reaching the
guarantee stability and to provide a good compromise boiling temperature of water and thus masking the
between accuracy and computation time. For solving Darcy's evaporation mechanism. Since estimability of
drying rate equations (ODEs), the Adams–Moulton's parameters α and β are in a grey zone, we tried to keep
predictor–corrector method was used since both Wp and their values constant, but it was found better to adjust
rc simultaneously vary with time. them.
Finally, the following subset of parameters (Wpc, Re,
4. Results and discussion α, β) was selected for the model calibration. Therefore,
parameters λp,w and KP had to be estimated independently,
4.1. Parameter estimability analysis either experimentally or using data from the literature. The
pellet permeability coefficient was determined experi-
The parameters were estimated separately for each set, mentally according to the procedure described by
A, B, and C, of drying conditions as presented in Table 1. Gelinas and Angers (1986). A value of KP = 3.28 · 10− 7 s
As discussed previously, the sets are obtained by coupling was obtained for that parameter and was kept constant
drying tests with weight and temperature monitoring in all simulations. The effective wet pellet thermal
corresponding to the same operating conditions (chamber
temperature and gas flowrate). Obviously the coupled Table 2
tests do not correspond to the same pellet. Since the pellet Estimability analysis results
mass differences in the same set are taken into account in Estimability order Parameter Residual sum of squares
the model, the drying phenomena are not synchronized in
1 Wpc 36.6200
both pellets, which mean for instance that the transition 2 Re 0.8300
from a drying mechanism to another one does not occur at 3 β 0.0770
the same time in the two pellets of a pair. 4 α 0.0473
The initial set of parameters to be estimated (Wpc, Re, 5 λp,w 0.0420
6 KP 0.0003
λp,w, α, β, KP) was submitted to a parameter estimability
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 105

conductivities λp,w was again estimated using Akiyama's contribute to increasing the water content at the pellet
approach (1991, 1992) as described in the Appendix. bottom. It is thus probable that the water film first
disappears at the pellet top part, and that the last wet part
4.2. Sharp transition between drying stages of the pellet is located below the pellet center.

The pellet drying model presented previously as- 4.3. Simultaneous surface and internal drying
sumes instantaneous transition from the film drying
stage to the shrinking core drying stage. Fig. 2 clearly To model such a transition period, one can model the
shows the difference between the smooth experimental evaporation mechanism using a two-dimensional de-
and the sharp transition predicted by the model. Surface scription (vertical coordinate and distance from the
drying still occurs in some region of the pellet surface, vertical axis) instead of the one-dimensional model used
while drying below the surface already occur in the here. However this would require a much more complex
remaining portion of the pellet. This is due to the in- set of equations which would be extremely difficult to
herent heterogeneities of the pellet properties as well as calibrate, using strategically and precisely placed ther-
of the environmental surrounding the pellets. Even in a mocouples. An empirical approach has been preferred
perfectly homogeneous environment dry spots would and another parameter, Wpc2 was introduced to define
appear randomly on the pellet surface and progressively the overall pellet moisture at the end of the transition
grow up until covering the whole surface. Here the period (beginning of the second drying phase), whereas
environment is also heterogeneous because the pellet Wpc defines the end of the first drying phase and the
lies on a piece of metal that probably extracts some heat beginning of the transition period. The drying rate
from the pellet. Furthermore, the hot gas is extracted during that transition period is assumed to be a linear
through a pipe placed just above the pellet. Due to this combination of the drying rates in the first and second
dissymmetry it is most probable that the evaporation is phases (as if these two phenomena would occur simul-
faster on the top of the pellet. In addition, gravity should taneously). The weight used in this linear combination

Fig. 2. Experimental and simulated mass and pellet temperatures under drying conditions of set A.
106 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

In the above criteria, simulated values are indicated by


the hat symbol. Also, subscript i is used to identify each
of the N measurement samples collected over time during
the drying tests. For each of the four measurements the
squared deviations are normalized by the squared experi-
mental values, thus assuming a constant relative standard
deviation of the measurement errors, except for the
surface temperature, for which the error standard devia-
tion is five times higher than for other measurements,
since it is uncertain whether the surface thermocouple
was really measuring the pellet surface temperature or the
Fig. 3. Variation of the Φ weighting factor of the evaporation rate.
gas temperature just above the surface.
Parameter estimation was performed independently
varies linearly between Wpc and Wpc2 as shown in for the three pairs A, B and C of drying tests in order to
Fig. 3, where Φ is the weight applied on the first phase verify whether parameter estimates depend on drying
drying rate, Wp is the pellet moisture content and Wp0 is conditions. A single set of parameters could also be
the initial pellet moisture content. estimated for all three pairs. Prior to the minimization of
Fig. 4 shows the behaviour of the temperatures ob- the criterion, the three experimental gas temperatures of
tained with the modified model compared to experimental each drying run were slightly corrected by simple trans-
results. Clearly the proposed approach allows to properly lations in such a way that their three asymptotic values
fit the transition, at the expense of having to calibrate an become equal to the average gas temperature. The
additional parameter. It is also satisfying that this empiri- estimated parameters are presented in Table 3. Criterion
cal approach well fit the linearly decreasing evaporation J3 correspond to the normalized and weighted sum of
rate that is observed during the transition period (see squares for the pellet drying rate calculated from mass
Fig. 4). Note that such a transition (i.e. heterogeneous versus time curves. It is shown here only as an indication
drying conditions) are also expected in industrial indura- since it was not used in the calibration criterion J.
tion machines since drying gas conditions are highly The critical humidity Wpc is reasonably constant for
variable through the bed (i.e channelling, etc.), and pellets the three test pairs. This is quite logical since this is an
are touching each other. This would create local variations intrinsic property of the pellet. The slight increase with the
in heat and mass transfer around the pellets and likely gas temperature and flow rate could be explained
yield a dissymmetrical drying front. If drying conditions considering that the first stage drying period is decreasing
were rather uniform within a pellet, the proposed model thus allowing a shorter time for the external water film to
would still be appropriate with Wpc2 =Wpc. be regenerated by water coming from the pellet core
The usual least squares method is used to estimate the through capillary forces (the duration of the first stage for
θ vector containing the five parameters of the model (Wpc, pairs A, B and C is respectively 1874, 592, and 357 s). The
Wpc2, Re, α, β). The calibration criterion is as follows: critical humidity at the end of the transition period is quite
significantly increasing as the evaporation rate increases
J ð hÞ ¼ J 1 þ J 2 ð12Þ due to the gas temperature and flow rate increase. This is
where J1 and J2 respectively correspond to the sum of the understandable since the duration of the transition period
squared deviations between the experimental values and is shorter (the duration of the transition for pairs A, B and
the simulated values for the pellet mass (m) and the three C is respectively 5412, 1862, and 406 s).
pellet temperatures (at the surface Ts, at the middle Tm, The effective diffusivity is the same for pairs A and
and at the center of the pellet Tc): B. Assuming that the pellet void fraction is around 25%,
the 4.5 value of the correction factor implies a tortuosity
XN  factor of about 1.1. This is a small value either indicating
mi  m̂i 2
J1 ¼ ð13Þ that the water diffusion in the pores is relatively easy
mi  0:01
i¼1 and/or that some convective transport is lumped in the
" #2 " #2 " #2 estimation of Dff. For pair C, the tortuosity is lower
X
N Ts;i  T̂ s;i Tm;i  T̂ m;i Tc;i  T̂ c;i (about 0.9). This implies that the effective diffusivity in
J2 ¼ þ þ the pores is close to the diffusion coefficient of water in
i¼1
Ts;i  0:05 Tm;i  0:01 Tc;i  0:01
free air. This might be due to an enlargement of the pores
ð14Þ produced by a fast evaporation rate, close to condition
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 107

Fig. 4. Simulated and experimental values for Test pair A.

where pellet spalling may occur, or again, to lumping of that the correlation used to predict km overestimate
convective transport. Note that values of β are of the the convective mass transfer. At the same time, in this
same order of magnitude compared to other porous correlation equation, the Reynolds number increases,
materials, such as catalysts, for which a typical range is such that there is probably some compensation between
4.0–15.0 (Villermaux, 1982). these two parameters and a negative statistical correla-
The Reynolds number increases as expected from the tion between their estimates. This problem cannot be
tests A to the tests C, due to the increase of the gas flow clarified because of the colinearity existing in the exper-
rate. The correction factor for convective mass transfer imental conditions of the three pairs of tests (the gas
α is decreasing from tests A to tests C, thus indicating flowrate and temperature are perfectly correlated).
108 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

Table 3
Estimates of the model parameters for each pair of drying tests
Test pair T∞
g , °C G, L/min Wpc, kg/m3 Wpc2, kg/m3 β (−) Re α J1 (N) J2 (N) J3
A 60 40 176 20.0 4.5 1100 0.596 1.3 (249) 39 (306) 979
B 100 60 183 38.9 4.6 1510 0.461 17.0 (121) 57 (161) 1400
C 150 80 222 138 2.5 2200 0.381 0.4 (61) 81 (81) 1193

Overall, the calibrated parameters seem to be com- presented in Figs. 4–6. Fig. 4 shows the pellet mass as
patible with the expected physical behaviour of pellet well as the three temperatures as functions of time. The
drying. One can look at the visual fitting between the fitting is excellent for the mass, the average standard
simulated and experimental data using the results deviation of the difference between the experimental

Fig. 5. Simulated and experimental values for Test pair B.


T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 109

Fig. 6. Simulated and experimental values for Test pair C.

and simulated masses being 0.12% of the average mass. probably not strictly spherical. The two figures showing
This value is comparable to the inherent accuracy of the the evaporation rate do not indicate any representation
thermo-balance. The temperatures are well represented problems. Obviously the standard error deviation is
by the model, although during the transition between the large (46%), but again compatible with the level of
two drying stages, the model still exhibits a plateau, uncertainty, which is large due to the noise amplification
which is not present in the data. The transition between introduced by the calculation of the derivative of the
the wet pellet and the fully dried pellet is not perfectly mass.
simulated, the temperature showing a smoother behav- For test pair B, while the temperatures do not
iour than predicted by the model. This is not surprising, exhibit biases, the model overestimates the pellet mass
since the end point of drying is probably not exactly reduction. This is corroborated by the overestimation
centered in the pellet but rather located in a region of the evaporation rate by about 20% during the drying
110 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

first stage period. Looking more carefully at the is the one proposed by Thurlby and Batterham (1980)
temperatures during this period shows a small and Thurlby et al. (1979):
overestimation of the temperature that might be
 
responsible for the overestimation of the evaporation 
dWp
dt
½
¼ vQconv = Cp;water Tb  Tp þ DH ðTb Þ
rate. Again, particularly at the pellet center, the behav-  
iour at the end of drying is too sharp in comparison to þCwv Tgl  Tb  ð15Þ
the experimental results. This is due to a combination
of various factors. First, as mentioned above, the where Qconv is the total amount of convective heat
drying end point cannot be as punctual as assumed by transferred (energy) by the gas to the pellet, and the
the model. Second, the presence of the thermocouple at denominator term within brackets represent the total
the pellet center disturbs the temperature profile. Third, amount of energy required for water evaporation. This
the numerical method used to solve the differential term consists of three parts: 1) the sensible heat required
equations makes use of too small control volumes in to raise the temperature of liquid water from the current
comparison to the wet core size close to the end of pellet temperature (Tp) to the boiling temperature (Tb), 2)
drying. the vapourisation energy of water at the boiling temper-
For the test pair C, the temperatures, mass and ature, and 3) the sensible heat required to raise the tem-
evaporation rates are reasonably well represented by the perature of water vapour from the boiling temperature to
model, considering that the experimental conditions are the current gas temperature (Tg). Adjustable parameter χ
quite different from the test pairs A and B. The drying therefore represents the fraction of the total amount of
time is 2000 s while it is 10 000 for A and 5000 for B. energy transferred by the gas to the pellet used for drying
The pellet mass is well fitted, although the evaporation specifically. According to Thurlby et al. (1979), the
rate is slightly underestimated. The temperatures are value of χ depends on drying conditions, but is usually
only slightly overestimated, and again there is a minor about 0.7.
misfit at the end of drying. A comparison between Thurlby's simplified second
In the two pellets the Darcy's evaporation regime phase drying expression (Eq. (15)) and the more com-
occurs respectively at times 1974 and 1511 s whereas plex expression used in this work, including the transi-
the core radius (rc) are 0.0052 and 0.00558 m (the pellet tion period between the first and the second drying
radius are 0.015 and 0.014 m). At this point, only 3.6% phases, is shown in Fig. 7. The drying rates are plotted
and 5.6% of the initial water content of the pellets is left against time and pellet moisture content. Fig. 7a) and b)
to evaporate. This explains why the permeability coeffi- corresponds to the measured drying rate obtained during
cient could not be estimated using the available experi- test 1, whereas Fig. 7c) and d), and e) and f) corresponds
mental data and had to be estimated separately since to drying rates obtained during tests 3 and 5 respective-
pellet pair C is the only set of experiments showing the ly. In all these figures, the blue dots and the red and the
Darcy's evaporation regime and that period is very black lines represent the measured experimental values,
short. To study this phenomenon in greater details, one the simulated drying rate using the expression devel-
needs to pursue drying experiments beyond the domain oped in this work, and the simulated drying rate using
investigated in this work using gas temperatures higher Thurlby's simplified expression respectively.
than 150 °C. For gas temperatures below 100 °C (Fig. 7a–d), there
is a very good agreement between Thurlby's expression
4.4. Comparison with simplified drying rate expressions and the one proposed in this work both at the beginning
and at the end of the drying curves. Note that for surface
Accounting for internal heat and mass transfer gra- drying Thurlby et al. (1979) used the same drying rate
dients in single pellet drying models requires some expression as the one used in this paper, and this
computing time to solve the various ordinary and partial explains the perfect agreement between the red and the
differential equations. However, simulating complete black curves early in the drying period. Since the values
moving grate furnaces involves drying of a pellet bed of the critical moisture content (Wpc) used by Thurlby
made of a very large number of single pellets. To reduce et al. (1979) are not specified in their paper, it was
computing time, ways to account for internal pellet decided to use those estimated in the presented study for
drying (i.e. second phase drying based on shrinking wet both approaches. The major discrepancy between the
core) without modelling internal heat and mass transfer two models occurs during the transition period between
gradients were proposed by a few groups. A good exam- the first and the second drying phases. Thurlby's ex-
ple of such a simplified second phase drying expression pression does not account for this phenomenon while
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 111

Fig. 7. Comparison between the second phase drying rate expression used in this work (red lines) and Thurlby's simplified expression (black lines). a
and b correspond to test 1, c and d to test 3, and e and f to test 5. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

the model proposed in this work fits the data very well in have enough flexibility to account for the complexity
this region. This comment also applies to all other of the experimental drying curves observed at this
previous works reviewed in this paper since, to the temperature.
author's knowledge, transition models were never
proposed for the drying of iron oxide pellets. For higher 5. Conclusion
gas temperatures (Fig. 7e–f) discrepancies between
Thurlby's model and the one proposed here are more The objective of this study was to assess the validity
important. It appears that Thurlby's expression does not of the usual assumptions about the drying mechanisms
112 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

of agglomerated iron oxide pellets. The existence of the N Number of data points
two distinct evaporation mechanisms of external water P atm External pressure atm
Pe Vapour pressure of water atm
film and shrinking core are confirmed. However, experi- Pr Prandtl number
mental results show that: Qconv Energy transferred to the pellet J/s
R Pellet radius m
– there is no sharp transition between the two mechanisms Re Reynolds number
r Distance from pellet center m
at a critical moisture content, but rather a smooth
rc Core radius m
transition period where the two mechanisms are active; Sc Schmidt number
– the moisture contents at the beginning and end of the Ta Initial temperature (ambient) K
transition period vary with the experimental conditions; Tb Water boiling temperature K
– the common assumption that a constant fraction of Tc Pellet temperature at the center K
the heat transferred to the pellet is used for water Tg∞ Gas temperature K
Tm Temperature at the pellet middle position K
evaporation must be replaced by a coupled heat and Tp pellet temperature K
mass balance model; Ts Pellet surface temperature K
– a third mechanism involving a bulk flow of water Vp Pellet volume m3
vapour from the wet core front to the pellet surface Wg∞ Gas humidity kg/ m3
appears when the temperature of the drying front is Wge Gas equilibrium humidity kg/ m3
Wp Pellet moisture content kg/m3
close to the boiling point. Wp0 Pellet initial moisture content kg/m3
Wpc, Wpc2 Critical pellet moisture contents kg/m3
Additional experimentation is required to identify the X Mass fraction of water in air on a dry basis kg water/
relationships between the moisture contents, at the kg dry air
beginning and end of the transition period, and the
Greek Letters
experimental conditions. Also, drying regimes at higher α Correction factor for convective mass transfer
temperature, when the wet core interface is above the β Correction factor for vapour diffusion
boiling point, thus eventually leading to pellet spalling, ξ Parameter in the pellet conductivity model
need to be investigated. ρg Gas density kg/m3
ρv Water vapour density kg/m3
ρp,d Dry pellet specific mass kg/m3
Symbols ρp,w Wet pellet specific mass kg/m3
a Specific pellet surface area m2/m3 μg Air viscosity kg/m s
per unit volume of pellet λFe2O3 Hematite thermal conductivity J/m s K
Acore Surface area of the wet core m2 λwv Water vapour thermal conductivity J/m s K
Ap Pellet surface area m2 λwl Liquid water thermal conductivity J/m s K
Ar Surface area at distance r from m2 λp,d Dry pellet thermal conductivity J/m s K
the pellet center λg Air thermal conductivity J/m s K
Cg Dry air heat capacity of the J/kg K λp,w Wet pellet thermal conductivity J/m s K
Cp,Fe2O3 Hematite heat capacity J/kg K ΔH Water evaporation enthalpy J/kg
Cp,SiO2 Silica heat capacity J/kg K Φ Weight of the film evaporation rate
Cp,w Wet pellet heat capacity J/kg K χ Fraction of energy used for water evaporation
Cp,d Dry pellet heat capacity J/kg K θ Set of parameters to be estimated or parameter
Cp,water Liquid water heat capacity J/kg K in the pellet conductivity model
Cwv Water vapour heat capacity J/kg K
dp Pellet diameter m
Dff Effective diffusion coefficient of m2/s Acknowledgement
water vapour in air
DH2O Diffusion coefficient of water vapour in air This work was financially supported by the Natural
G Superficial gas flow rate kg/m2 s
H Convective heat transfer coefficient J/m2 s K
Sciences and Engineering Research Council of Canada
i Measurement index (NSERC).
J, J1, J2, J3 Least squares criteria
km Convective mass transfer coefficient m/s Appendix A
at the pellet surface
Kp Relative permeability S
kpar, kser, Thermal conductivities used in J/m s K
A.1. Gas properties
k1, k2 Akiyama's expression
m Pellet mass kg The following properties of the drying gas (air)
Mp,0 Pellet initial mass kg are necessary for the evaluation of convective heat
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 113

and mass transfer coefficients (all temperatures are The first was estimated using the drying data whereas the
in K). last two are defined as follows:
Thermal conductivity (Küçükada et al., 1994) (J/m s K):
Pr ¼ lg Cg =kg ðA2:3Þ
2:11  103 Tg1:5
kg ¼ ðA1:1Þ
Tg þ 123:6 Sc ¼ lg =qg DH2 O ðA2:4Þ

Gas density (Küçükada et al., 1994) (kg/m3): The diffusion coefficient of water vapour in air is given
by the following expression (Treybal, 1980) (m2/s):
219:38ð1 þ X Þ
qg ¼ ðA1:2Þ  3=2
ð X þ 0:622ÞTg Tg
DH2 O ¼ 2:58e5 : ðA2:5Þ
where X is the mass fraction of water in air on a dry 300
basis.
Gas viscosity (Rohsenow and Harnett, 1973) (kg/m s): A.3. Pellet properties

lg ¼ 107 ð4:02 þ 7:46e1 Tg  5:72e4 Tg2 Heat capacity of the dry and wet pellet (Caron and Roy,
7 11 1996) (J/kg K):
þ 2:99e Tg3  6:25e Tg4 Þ ðA1:3Þ
The heat capacity of the dry pellet was assumed to be
Heat capacity (Smith and Van Ness, 1987) (J/kg K): close to that of the iron oxide concentrate, which is
roughly composed of 95% Fe (taken as Fe2O3) and 5%
Cg ¼ Cp;dry air þ Cwv X ðA1:4Þ silica:
where Cp;d ¼ 0:95  Cp;Fe2 O3 þ 0:05  Cp;SiO2 ðA3:1Þ
!
8314 0:02e5 where
Cp;dry air ¼ 3:36 þ 0:58e3 Tg 
29 Tg2   " #
4187 3 1:44  105
ðA1:5Þ Cp;SiO2 ¼  10:49 þ 0:24  10 Tp 
60 Tp2
! ðA3:2Þ
8314 0:12e5
Cwv ¼ 3:47 þ 1:45e3 Tg þ ðA1:6Þ
18 Tg2 9:23  106
Cp;Fe2 O3 ¼ 614:260 þ 0:486Tp  : ðA3:3Þ
Tp2
A.2. Convective heat and mass transfer coefficients
When the pellet is wet, the heat capacity is computed
The following heat transfer correlation developed for as a weighted average of the heat capacity of the dry
single sphere is used to estimate the heat transfer coef- pellet and that of water:
ficient (h) (Seshadri and da Silva Pereira, 1985;
Küçükada et al., 1994; Geankoplis, 1978) (J/m2 s K): Cp;d  qp;d þ Cp;water  Wp
Wp NWpc Cp;w ¼ ðA3:4Þ
qp;d þ Wp
kg  
h¼ 2 þ 0:6 Re1=2 Pr1=3 : ðA2:1Þ
dp Cp;d  qp;d þ Cp;water  Wpc
Wp bWpc Cp;w ¼ ðA3:5Þ
The convective mass transfer coefficient (km) is ob- qp;d þ Wpc
tained using the Chilton–Colburn analogy (Seshadri and where
da Silva Pereira, 1985; Küçükada et al., 1994; Geanko- !
plis, 1978) (m/s): 4184 0:08  105
Cp;water ¼ 7:17 þ 2:56  103 Tp þ
DH2 O   18 Tp2
km ¼ 2 þ 0:6  Re1=2  Sc1=3  a ðA2:2Þ
dp ðA3:6Þ
where α is a correction factor (estimated from the data), and
dp is the pellet diameter, Re, Pr, and Sc are the Reynolds,
the Prandtl, and the Schmidt dimensionless numbers. qp;w ¼ qp;d þ Wp ðA3:7Þ
114 T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115

The density of the dry pellet ρp,d was calculated A very good fit of the pellet temperatures was ob-
using pellet mass and volume. tained (see Tsukerman (2006) for more details on this
Thermal conductivity of the dry and wet pellet (J/s experiment).
m K): Effective diffusion coefficient:
The approach proposed by Akiyama et al. (1991, The effective diffusion coefficient of water vapour
1992) was selected in this work. It assumes that the within the dry shell of the pellet was obtained using the
porous material can be decomposed in elementary cells following expression (Hasenack et al., 1975; Seshadri
with a core part and connecting parts arranged in some and da Silva Pereira, 1985; Chen and Pei, 1989;
way, which yields a pore distribution model. The effec- Küçükada et al., 1994; Davis, 1998):
tive thermal conductivity of such a material is computed
as the geometric mean of the following two thermal DH2 O
Dff ¼ ðA3:13Þ
conductivities, kser and kpar, representing the resistance b
in series of the pore distribution model (minimum
value), and the resistance in parallel of the same model where DH2 O is calculated using Eq. (A2.5) and β an
(maximum value): estimated parameter. It corresponds to the ratio of pore

  31 tortuosity to pellet porosity.
2
nh= k1 þ ðk2  k1 Þ 1  h2  2nhð1  hÞ

 
6 Þ 1  h2 o 7
kser ¼ 4 þðh  nhÞ=n k1 þ ðk2  k1 5 A.4. Other thermodynamic properties
2
þð1  hÞ= k1 þ ðk2  k1 Þ 1  ðnhÞ
Heat evaporation of water (Küçükada et al., 1994) (J/kg):
ðA3:8Þ
DH ¼ 3:16  106  2495:45Tp þ 0:29Tp2 : ðA3:14Þ

kpar ¼ k2 1 "2nhð1  hÞ  h2 þ k1 ðnhÞ2 # Equilibrium water vapour pressure (Reid et al., 1987)
n ð1  hÞ=fko1 ð1  nhÞ þ k2 nhg
2nh (kPa):
þk2 k1  
þ h2  ðnhÞ2 =fk1 ð1  hÞ þ k2 hg
1
Pe ¼ exp  ð7:765  x þ 1:458  x1:5
ðA3:9Þ 1x
where θ and ξθ are structural parameters characterizing  2:776  x3  1:233  x6 Þ  Pc  100
the core and connecting parts of the elementary cells ðA3:15Þ
making the porous material, and k1 and k2 are the con- T
ductivities of the solid phase (here hematite) and the pore where x ¼ 1  Tpc , and Tc and Pc are the critical temper-
phase, made of either air, liquid water, or vapour. The air ature and pressure of water. The gas moisture content at
thermal conductivity is calculated using Eq. (A1.1) saturation can therefore be computed as follows, as-
whereas the conductivity of hematite (Akiyama et al., suming ideal gas law is valid:
1992), liquid water and water vapour (Reid et al., 1987) Patm  18
are obtained using the following expressions: Wge jr¼rc ¼ ðA3:16Þ
R  Tp ðrc ;t Þ
1
kFe2 O3 ¼ ðW=m−KÞ ðA3:10Þ
1:844e4 Tp References

kwl ¼ 0:38 þ 5:25e3 Tp  6:37e6 Tp2 ðA3:11Þ Akiyama, T., Takahashi, R., Yagi, J., 1991. Heat transfer simulation on
drying processes of nonfired pellets containing combined water in
the moving bed reactor. ISIJ Int. 31, 24–31.
kwv ¼ 7:34e3  1:01e5 Tp þ 1:80e7 Tp2 Akiyama, T., Ohta, H., Takahashi, R., Waseda, Y., Yagi, J., 1992.
 9:10e11 Tp3 : ðA3:12Þ Measurement and modeling of thermal conductivity for dense iron
oxide and porous iron ore agglomerates in stepwise reduction. ISIJ
For nonfired hematite pellets having a porosity in the Int. 32, 829–837.
order of 25%, as is the case in this study, Akiyama et al. Breiholtz, C., Hillberg, C., 1980. Development of a reference model for
the drying zone of a travelling grate pelletizing plant. IFAC Mining,
(1992) suggest to use structural parameter values of
Mineral and Metal Processing. Montréal, Canada, pp. 415–424.
θ =0.906 and ξθ = 0.150. This model was validated Caron, S., Roy, D., 1996. SIMBOUL intégré V1.0: Simulateur du
against data obtained for heat transfer in dry pellets procédé de cuisson des boulettes de concentré de fer. Project Report
using the same experimental setup as for drying tests. 7234 P 026. COREM, Québec, Canada.
T. Tsukerman et al. / Int. J. Miner. Process. 83 (2007) 99–115 115

Chen, P., Pei, D.C.T., 1989. A mathematical model of drying Nekrasova, E.V., Butkarev, A.P., Maizel, G.M., Klein, V.I., Kuznetsov,
processes. Int. J. Heat Mass Transfer 32, 297–310. R.F., 1988. Drying mechanism of iron ore pellets. Steel USSR 18,
Clark, K.N., 1981. Iron ore pellet drying mechanisms under the 249–251.
heating conditions encountered in a straight-grate indurator. Trans. Patisson, F., Bellot, J.P., Ablitzer, D., 1990. Study of moisture transfer
Inst. Min. Metall. C. 90C, C66–C72. during the strand sintering process. Metall. Trans., B, Process
Cross, M., Wade, K.C., 1989. Computer simulation of iron ore pellet Metall. 21B, 37–47.
induration with additives. In: Cross, M., Oliver, R. (Eds.), Proc. 5th Pomerleau, D., Hodouin, D., Poulin, É., 2005. A first principle
Int. Symp. on Agglomeration. IChemE, London, UK, pp. 291–298. simulator of an iron oxide pellet induration furnace — an
Cross, M., Blot, P., 1999. Optimizing the operation of straight-grate application to optimal tuning. Can. Metall. Q. 44, 571–582.
iron-ore pellet induration systems using process models. Metall. Reid, R.C., Prausnitz, J.M., Poling, B.E., 1987. The Properties of
Mater. Trans., B, Proc. Metall. Mater. Proc. Sci. 30B, 803–813. Gases and Liquids, fourth ed. McGraw-Hill, New-York.
Cross, M., Gibson, R.D., Young, R.W., 1979. Pressure generation Rohsenow, W.M., Harnett, J.P., 1973. Handbook of Heat Transfer.
during the drying of a porous half-space. Int. J. Heat Mass Transfer McGraw-Hill, New York.
22, 47–50. Seshadri, V., da Silva Pereira, R.O., 1985. Mathematical simulation of
Cumming, M.J., Thurlby, J.A., 1990. Developments in modelling and induration of iron ore pellets in pot grate. Proc. 4th Int. Symp. on
simulation of iron ore sintering. Ironmak. Steelmak. 17, 245–254. Agglomeration. ISS-AIME, Toronto, Canada, pp. 729–744.
Cumming, M.J., Rankin, W.J., Siemon, J.R., Thurlby, J.A., Thornton, G.J., Smith, J.M., Van Ness, H.C., 1987. Introduction to chemical
Kowalczyk, E.A., Batterham, R.J., 1985. Modelling and simulation of engineering thermodynamics. McGraw-Hill, New York.
iron ore sintering. Proc. 4th Int. Symp. on Agglomeration. ISS-AIME, Thurlby, J.A., 1988. A dynamic mathematical model of the complete
Toronto, Canada, pp. 763–776. grate/kiln iron-ore pellet induration process. Metall. Trans., B,
Dash, I., Carter, C.E., Rose, E., 1974, Heat wave propagation through Process Metall. 19B, 103–111.
a sinter bed; a critical appraisal of mathematical representations. In: Thurlby, J.A., Batterham, R.J., 1980. Prediction of drying and spalling
SIMAC 74, Harrogate, Paper No. 8, 1974, P8/1–7. behaviour of hematite pellets. Trans. Inst. Min. Metall. C. 89C,
Davis, R.A.J., 1998. Oxide formation in an iron oxide pellet rotary kiln C125–C131.
furnace. Air Waste Manag. Assoc. 48, 44–51. Thurlby, J.A., Batterham, R.J., Turner, R.E., 1979. Development and
Geankoplis, C.J., 1978. Transport Processes and Unit Operations. validation of a mathematical model for the moving grate induration
Allyn and Bacon, Boston. of iron ore pellets. Int. J. Miner. Process. 6, 43–64.
Gelinas, C., Angers, R., 1986. Improvement of the dynamic water- Treybal, R.E., 1980. Mass Transfer Operations. McGraw-Hill,
expulsion method for pore size distribution measurements. Am. New-York.
Ceram. Soc. Bull. 65, 1297–1300. Tsukerman, T., 2006. Modelling and simulation of the drying kinetics
Hasenack, N.A., Lebelle, P.A.M., Kooy, J.J., 1975. Induration process of iron oxide pellets. M.Sc. Thesis, Department of Chemical
for pellets on a moving strand. Mathematical Process Models in Engineering, Université Laval, Québec, Canada.
Iron- and Steel-Making. TMS, Warrendale, PA, pp. 6–16. Villermaux, J., 1982. Génie de la réaction chimique: conception et
Ilic, M., Turner, I.W., 1986. Drying of a wet porous material. Appl. fonctionnement des réacteurs. Lavoisier, Paris.
Math. Model. 10, 16–24. Voskamp, J.H., Brasz, J., 1975. Digital simulation of the steady-state
Küçükada, K., Caron, S., Hodouin, D., Ouellet, G., Paquet, G., behaviour of moving bed processes. Meas. Control 8, 23–32.
Thibault, J., 1992. Mathematical modelling of the iron ore pellet Young, R.W., Cross, M., Gibson, R.D., 1979. Mathematical model of
induration process. 75th Steelmaking, 51st Ironmaking and 10th grate-kiln-cooler process used for induration of iron ore pellets.
Process Technology Division Conferences, Toronto, Ontario, Ironmak. Steelmak. 6, 1–13.
Canada, April 5–8. Zhen, Y., Shaw, K., Kou, B.M., McAuley, K.B., Bacon, D.W., 2003.
Küçükada, K., Thibault, J., Hodouin, D., Paquet, G., Caron, S., 1994. Modeling ethylene/butene copolymerization with multi-site cata-
Modelling of a pilot scale iron ore pellet induration furnace. Can. lysts: parameter estimability and experimental design. Polym.
Metall. Q. 33, 1–12. React. Eng. 11, 563–588.
Muchi, I., Higuchi, J., 1972. Theoretical analysis of sintering
operation. Trans. Iron Steel Inst. Jpn. 12, 54–63.

You might also like