Drying Technology

An International Journal

ISSN: 0737-3937 (Print) 1532-2300 (Online) Journal homepage: https://www.tandfonline.com/loi/ldrt20

Numerical Spray Model of the Fluidized Bed

Coating Process

F. Ronsse , J. G. Pieters & K. Dewettinck

To cite this article: F. Ronsse , J. G. Pieters & K. Dewettinck (2007) Numerical Spray
Model of the Fluidized Bed Coating Process, Drying Technology, 25:9, 1491-1514, DOI:
10.1080/07373930701537245

To link to this article: https://doi.org/10.1080/07373930701537245

Published online: 10 Sep 2007.

Submit your article to this journal

Article views: 254

View related articles

Citing articles: 5 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=ldrt20
Drying Technology, 25: 1491–1514, 2007
Copyright # 2007 Taylor & Francis Group, LLC
ISSN: 0737-3937 print/1532-2300 online
DOI: 10.1080/07373930701537245

Numerical Spray Model of the Fluidized Bed Coating Process

F. Ronsse,1,2 J. G. Pieters,1 and K. Dewettinck2
1
Biosystems Engineering, Ghent University, Gent, Belgium
2
Food Technology and Engineering, Ghent University, Gent, Belgium

more downward-spraying nozzles are usually located

In this study, a new model for the batch top-spray fluidized above the bed.[4,5] The use of compressed air in a binary
bed coating process is presented. The model is based on the one-dimen- nozzle results in very strong shear forces at the liquid-gas
sional (axial) discretization of the bed volume into different control interface, producing droplets with a size ranging from 10
volumes, in which the dynamic heat and mass balances for air, water to 40 mm.[5,6] The fluidizing air also supplies the evaporative
vapor, droplets, core particles, and coating material were established.
The coupling of the droplet phase’s mass and heat transfer terms with capacity to remove the solvent, leaving behind the dis-
the gas and solid phases was established by means of a droplet submo- solved material on the surface of the fluidized particles.
del in which droplet trajectories were individually simulated. When liquid binders or dissolved coating polymers are
The model calculation method combines a Monte Carlo tech- added to a fluidized bed, two different particle growth
nique for the simulation of the particle exchange with the first-order modes are to be distinguished, depending on the mass rate
Euler’s method for solving the heat and mass balances, enabling the
prediction of both the dynamic coating mass distribution and the of binder addition and the evaporative capacity of the
one-dimensional (axial) thermodynamic behavior of the fluidized bed.[7] When wetted particles collide, a wet bridge is formed
bed during batch operation. The simulation results were validated in between them. Depending on the liquid bridge strength
using experimental two-dimensional spatial air temperature and air and the kinetic energy of the colliding particles, these
humidity distributions, which were measured in a fluidized bed pilot bridges may persist beyond the point of solidification or
reactor using a scanning probe.
Sensitivity analysis was carried out to study the effect of control- drying and, consequently, dry agglomerates or clusters
lable process variables, such as fluidization air and atomization air are formed.[8–10] This type of particle growth is termed
properties, as well as the properties of the spraying liquid upon the ‘‘agglomeration.’’ When fluidized beds operate in dryer
simulated dynamic temperature and humidity distributions. Also, conditions; i.e., when the evaporative capacity of the bed
the effects of relevant process variables on growth rate uniformity is increased or when the rate at which the liquid binder is
and process yield were studied. Based on these sensitivity studies
it was concluded that nozzle parameters, such as air pressure and added to the bed is reduced, the collected coating solution
positioning with respect to the bed, are as important as the fluidiza- on the particle surface will sufficiently dry before collision
tion air properties (humidity, temperature, and flow rate) for the and, consequently, the formation of liquid bridges is
coating growth rate uniformity and process yield. prevented.[1] Previous studies carried out by Smith and
Nienow[11] and by Maronga and Wnukowski[12,13] have
Keywords Coating mass distribution; Fluidized bed coating; shown that particles are uniformly wetted in a small region
Heat and mass transfer; Modeling; Spray drying in the proximity of the spraying nozzle. This so-called wet-
ting zone corresponds to the droplet penetration depth of
INTRODUCTION the spray. The repeated cycle of wetting and drying when
Fluidized bed coating can be defined as an encapsula- the fluidized particles move from and toward the wetting
tion or surface modification process of particulate solids zone results in the creation of a uniform and layered struc-
by which a coating polymer is applied into a fluidized ture around the individual particles. This second type of
bed. The coating polymer to be applied could be an aque- particle growth is also termed ‘‘layering.’’[1]
ous or organic solvent–based solution or even a melt and is The only type of growth mechanism that is preferred in
continuously sprayed, usually by means of a pneumatic or coating applications is layered growth. In order to promote
binary nozzle, which may be submerged in or positioned layered growth and to suppress agglomeration, dry process
above the bed.[1–3] In the top-spray configuration, one or conditions are required. However, under such dry con-
ditions and considering that the coating solution is sprayed
Correspondence: F. Ronsse, Biosystems Engineering, Ghent countercurrently with the fluidizing air, premature droplet
University, Coupure Links 653, B-9000 Gent, Belgium; evaporation is likely to occur before the coating solution
E-mail: [email protected]

1491
1492 RONSSE ET AL.

droplet is collected onto the particle surface.[14,15] The depending on the thermodynamic characteristics of the
resulting spray-dried coating material is partially collected fluidized bed. This model could serve, in future studies,
by the filter system, partially entrapped within the coating as a basis for process control or reactor design in fluidized
film, resulting in coating imperfections.[8] bed coaters to further increase coating efficiency and coat-
Currently, the design and optimization of fluidized bed ing uniformity or to reduce energy consumption.
coating processes is not possible without extensive trial-
and-error testing due to the large amount of variables MODEL DESCRIPTION
involved.[1,16] Process models prove to be useful tools in
Fluidized Bed Coater Model
understanding or clarifying the impact of the different
Introduction
input variables on process efficiency and could therefore
substantially reduce the research and design work required In order to develop the model, the fluidized bed was
for the successful coating of a new product in a fluidized one-dimensionally discretized into n control volumes, each
bed or for the implementation of new reactor designs or having a constant volume of Vbed =n (Fig. 1) and a constant
improved process control methods.[17] Currently, little is number of particles Ni. Within each control volume, three
known regarding the interactions between the different different phases are distinguished: the gas (air) phase, the
microprocesses involved in fluidized bed coating, such as solid phase (particles), and the droplet phase (Fig. 2). In
droplet production, evaporation, heat transfer, droplet modeling these three phases within the fluid bed, the fol-
impingement, and particle behavior in a fluidized bed lowing assumptions were made:
and how these interactions relate to process yields, coating
Each phase (gas, solids and droplets) is perfectly
mass distributions, and coating uniformity. mixed within each control volume Si .
Therefore, the aim of this study is the development of a
The mass flow rate of fluidization air (Ga,i) is con-
mass and heat transfer model for the description of the stant for each control volume. Atomization air
dynamic temperature and concentration distributions of (Gat) is assumed to be uniformly mixed over the
liquid-sprayed fluidized beds. Also, the model aims at upper c control volumes. The mixed air (fluidization
predicting the coating mass distribution and the extent air and atomization air) is assumed to move in a plug
to which premature droplet evaporation takes place flow regime between the different control volumes.

FIG. 1. The top-spray batch fluidized bed reactor and the discretization of the bed into different compartments or control volumes.
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1493

FIG. 2. Detail of a modeled control volume and the interactions between the different phases inside a modeled control volume.

Particles in each control volume Si are continu-
Droplets travel downward through the particle
ously being exchanged with the neighboring con- bed. Before contacting the particle surface, dro-
trol volumes Si1 and Siþ1. When applying the plets exchange heat and mass with the surrounding
model to monodisperse particle populations, the air. If no successful adhesion occurs before com-
exchange probability is considered to be equal plete droplet evaporation, dry fines are produced.
for each particle in Si. Particle exchange is In this model, dry fines are assumed to be com-
expressed by ri, as the fraction of the particle pletely removed from the bed by the fluidizing
population exchanged per time unit through the air. In reality, however, part of the spray-dried
interface between control volumes Si and Siþ1. fines may stick to the surface of a wetted particle,

The particles are mechanically inert; there is resulting in coating imperfections.[8]
neither agglomeration nor attrition. Particles are
considered to be nonporous (e.g., glass beads)
and do not absorb coating solution. Due to the Heat and Mass Balances
low level of liquid loading of the fluidized bed, In each control volume the dynamic heat and mass
characteristic of coating processes, the impact of balances for air, particles, and coating material were estab-
liquid bridge formation on the fluidization lished and the following equations were obtained:
behavior—and hence, the particle exchange
rate—is considered not to be significant. Accord- Particle Balance in a Single Control Volume. Resulting
ing to Schaafsma et al.,[18] particle mixing behavior from the third assumption, the particle population balance
and minimum fluidization gas velocity are not sig- could be written as:
nificantly influenced by liquid bridge formation as dNi
long as the relative humidity of the bed’s gas phase ¼ ðri1 Nbed þ ri Nbed Þ  ðri Nbed þ ri1 Nbed Þ ¼ 0 ð1Þ
dt
does not exceed 0.5.

Since the model was designed to handle coating
processes with small film thicknesses (micrometer
Moisture Balance of the Particles. The moisture
range), the weight of the added coating to the par-
balance in each control volume Si is governed by the
ticles is relatively small compared to the weight of
supply and removal of wetted particles by exchange from
the particles themselves. Consequently, the weight
and toward adjacent control volumes, condensation or
of the individual particle is assumed to be constant
evaporation at the particle surface, and droplet collection:
throughout the coating process.

The outlet is at atmospheric pressure. The pressure dWp;i
drop across the fluidized bed is small compared to Mp Ni ¼ Mp Nbed ri1 Wp;i1 þ Mp Nbed ri Wp;iþ1
dt
the overall atmospheric pressure. Consequently,
 Mp Nbed ðri þ ri1 ÞWp;i  RD;i Ni Mp
the drying process is assumed to take place at con-
stant atmospheric pressure. þ RC;i Ni Mp ð1  DMdr;i Þ ð2Þ
1494 RONSSE ET AL.

Moisture Balance of the Droplets. The droplet moisture Heat Balance of the Particles. Considering the diameter
balance in a control volume Si is composed of moisture in- range of both droplets (0  ddr  50 mm) and particles
and output from the moving droplet phase, evaporation (100 mm  ddp  500 mm), it holds that the Biot number:
and=or condensation at the droplet surface, and droplet
ap dp adr ddr
collection onto the particle surface. It is possible that a Bip ¼ << 0:1 and Bidr ¼ << 0:1 ð7Þ
fraction of the droplets does not impinge on the particle kp kdr
surface but on the inner walls of the reactor; hence, the Consequently, the internal heat transfer resistance is
variable R1C;i in the collection term of Eq. (3): negligible compared to the convective heat transfer
d½Mdr;i ð1  DMdr;i Þ resistance at the particle or droplet surface. Therefore,
¼ ð1  DMdr;iþ1 ÞJdr;iþ1 each droplet or particle is considered to be isothermal and
dt
could be represented as having a single temperature.[19,20]
 ð1  DMdr;i ÞJdr;i  RD;i Mdr;i
The particle heat balance of the control volume Si takes
 ðRC;i Ni Mp þ R1
C;i Awall;i Þð1  DMdr;i Þ into account particle transport from and toward adjacent
ð3Þ control volumes (Si1, Siþ1), convective heat exchange
with fluidizing air, and the heat required for the collected
droplets to reach particle temperature, Tp,i:
Moisture Balance in the Gas (Air) Phase. The change in
air moisture in a control volume Si is determined by the dTp;i
Ni Mp Cp;p ¼ Mp Nbed ri Cp;p Tp;iþ1 þ Mp Nbed ri1 Cp;p Tp;i1
moisture in the incoming process air from Si1, the moisture dt
in the process air leaving from Si to Siþ1, and evaporation  Mp Nbed Cp;p Tp;i ðri þ ri1 Þ
or condensation on the surface of both particles and þ ap;i Ni Ap ðTa;i  Tp;i Þ  RD;i Ni Mp Qlat;i
droplets. The amount of compressed air that is released to
assist in atomizing the coating solution was assumed to be þ RC;i Ni Mp Cp;dr;i ðTdr;i  Tp;i Þ  gp;i Uloss;i
homogeneously divided over the upper c control volumes ð8Þ
and, therefore, the air moisture balance could be written as:
In Eq. (8) is Uloss,i, the heat loss term in the control volume
dXa;i j Si, which can be splitted into two parts: particles-to-
Ma;i ¼ Ga;i1 Xa;i1  Ga;i Xa;i þ Gat Xat
dt c environment heat loss due to the occurrence of combined
þ RD;i Ni Mp þ R1 
D;i Awall;i þ RD;i Mdr;i convection and conduction at particle-wall contact and

fluidizing air-to-environment heat loss due to convection.
i<c)j¼0
with Hence the term gp,i, which expresses the fraction of the
cin)j¼1 ð4Þ
total heat loss in Si originating from the particle phase.
Heat Balance of the Droplets. Combining the enthalpy
Coating Mass Balance of the Particles. Analogous to of the droplet phase input and droplet phase output in Si,
the moisture balance, except for the drying term, the convective heat transfer between droplet phase and gas
coating mass balance could be written as: phase (air), evaporation at the droplet surface with the
dYp;i different droplet sink terms—such as droplet collection
Mp N i ¼ Mp Nbed ri1 Yp;i1 þ Mp Nbed ri Yp;iþ1 onto particles or reactor wall and the removal of spray-
dt
dried fines—yields the following droplet heat balance:
 Mp Nbed ðri þ ri1 ÞYp;i þ RC;i Ni Mp DMdr;i
ð5Þ dTdr;i
Mdr;i Cp;dr;i ¼ Jdr;iþ1 Cp;dr;iþ1 Tdr;iþ1  Jdr;i Cp;dr;i Tdr;i
dt
Coating Mass Balance of the Droplets. Depending on  ðRC;i Ni Mp þ R1 C;i Awall;i ÞCp;dr;i Tdr;i
the droplet drying rate and the diameter of the droplets  Jsd;i Cp;c;i Tdr;i  RD;i Mdr;i Qlat;i
in each control volume Si, a fraction of the droplets will
be spray dried, resulting in the formation of dry fines. It þ adr;i Ni Adr;i ðTa;i  Tdr;i Þ ð9Þ
was assumed that dry fines are entirely removed from the In Eq. (9) is Ni , the number of droplets contained in the
bed by the fluidizing air; therefore, the coating mass control volume Si. The initial number of droplets, Nn , is
balance of the droplet phase could be written as: based on the liquid feed rate Jsol, the initial droplet size
dðMdr;i DMdr;i Þ dd,0, and the droplet nozzle exit velocity, vd,0. The
¼ DMdr;iþ1 Jdr;iþ1  DMdr;i Jdr;i number of droplets, Ni , becomes smaller as the droplet
dt
 ðRC;i Ni Mp þ R1 phase migrates to the bottom of the fluidized bed. This
C;i Awall;i ÞDMdr;i  Jsd;i
number of droplets, Ni , is determined by using the
ð6Þ following population balance, describing droplet number
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1495

decrease due to spray drying and droplet collection onto

the particle or onto the reactor wall:
dNi   6Jsd;i
¼ Ndr;iþ1  Ndr;i  3
dt pqc dsd;i
6
 3
ðRC;i Ni Mp þ R1
C;i Awall;i Þ ð10Þ
pqdr;i ddr;i

Heat Balance of the Gas (Air) Phase. The equation for

the enthalpy balance of the gas phase (air) within each
control volume Si is given by the enthalpy of the air
entering Si from Si1, the enthalpy of the air leaving Si
to Siþ1, the enthalpy of the supplied atomization air, the
heat transferred between the gas phase and the droplet
phase, the heat transferred between the gas phase and the
particles in Si, and the heat losses toward the environment:
dTa;i
Ma;i Cp;a;i ¼ Ga;i1 Cp;a;i1 Ta;i1  Ga;i Cp;a;i Ta;i
dt
j
þ Gat Cp;at Tat  ð1  gp;i ÞUloss;i
c
 ap;i Ni Ap ðTa;i  Tp;i Þ
 adr;i Ni Adr;i ðTa;i  Tdr;i Þ
 RD;i Ni Mdr;i Cp;v;i ðTa;i  Tdr;i Þ
 RD;i Ni Mp Cp;v;i ðTa;i  Tp;i Þ
 R1
D;i Awall;i Cp;v;i ðTa;i  Twall;i Þ
FIG. 3. Detail of the modeled reactor wall element with the different

local heat transfer phenomena.
i<c)j¼0
with
cin)j¼1 (11) particle phase and the inner reactor wall was neglected,
as this mode of heat transfer is not significant when the
fluidized bed operates at temperatures below 400 K.
Heat Loss and Wall Element Second, the gas phase—more specifically, the bubble gas
To quantify the overall heat losses in each control phase—sweeps across the inner reactor, resulting in the
volume Si, the cylindrical reactor shell was modeled into convective heat transfer, expressed by the transfer
different elements or control volumes (Fig. 3); the number rate ð1  gp;i ÞUloss;i . For a more extensive overview and
of wall elements equaled the number of fluidized bed con- quantification of the different heat transfer mecha-
trol volumes. Due to the relatively small reactor wall thick- nisms in fluidized beds, the reader is referred to Kunii
ness compared to the reactor wall height (dwall << hwall ), and Levenspiel.[22]
only axial conduction (Ucond,i) was to be considered as heat Finally, depending on the thermodynamic operation
transfer mechanism inside the reactor shell of the wall point of the fluidized bed and the nozzle and coating sol-
control volume, Swall,i. ution parameters influencing the initial droplet size distri-
At the outside of the wall element, radiation (Urad,i) and bution (dd,0) and droplet exit velocity (vd,0), a fraction of
natural convection (Uconv,i) were considered. The convec- the produced droplets will impact on the inner reactor wall.
tive heat transfer coefficient (awall) was approximated using The heat required to evaporate the wall-collected droplet
the dimensionless Nusselt numbers for natural convection phase has to be substracted from the wall element heat
around a free-standing vertical pipe.[21] The heat transfers balance (Fig. 3). Consequently, the overall heat balance
between the inner reactor wall and the fluidized were split of a single wall element Swall,i could be written as:
into two parts. First, particles within the bed exchange heat
with the inner wall through the combination of the dTwall;i
wall-particle contact point (conduction) and the thin gas dwall;i Awall;i qwall Cp;wall ¼
dt
layer surrounding the wall-particle contact point (convec-
Uloss;i þ Ucond;i1  Ucond;i  Uconv;i
tion and conduction), expressed by the heat transfer rate
gp;i Uloss;i . Heat transfer through radiation between the  Urad;i  Uevap;i þ Udr;i ð12Þ
1496 RONSSE ET AL.

kwall Droplet Model

¼ Uloss;i þ p dwall;i Ddwall ðTwall;i1  Twall;i Þ
hwall;i Introduction and Model Assumptions
kwall In the model presented in the previous section, the drop-
p dwall;iþ1 Ddwall ðTwall;i  Twall;iþ1 Þ let collection rates ðRC and R1
hwall;iþ1 C Þ and the droplet drying
  rate ðRD Þ in each control volume are still unknown. Also,
 awall Awall;i ðTwall;i  Te Þ  r e0wall Awall;i Twall;i
4
 Te4 the degree to which droplets evaporate (Jsd) resulting in
 R1 1 the formation of dry fines as a function of the control
D;i Awall;i Qlat;i þ RC;i Awall;i Cp;dr;i ðTdr;i  Twall;i Þ
volume’s air thermodynamic properties is unknown. In
ð13Þ order to quantify these unknown variables, a separate
droplet submodel was coupled with the main model. The
Gas=Solid Heat and Mass Transfer Rates aim of this droplet submodel is to calculate the spatial dis-
In the above equations, besides the droplet-related tribution of the sprayed liquid inside the fluidized bed, the
terms, the heat and mass transfer rates are yet unknown. rate at which droplets adhere onto the fluidized particles
The convective heat transfer coefficient between the parti- and the solvent evaporation occurring between droplet
cles (ap) and the gas (air) phase (ac) was approximated production at the nozzle and droplet collection onto the
using the dimensionless Nusselt numbers. The Nusselt particle surface.
number for forced convection around spherical bodies In this submodel, the droplet behavior is determined by
was calculated using the Whitaker equation:[23] tracking the trajectories of individual droplets throughout
!1=4 the computational domain. The following assumptions
1=2 2=3 2=5 la;i were made:
Nup;i ¼ 2 þ ð0:43 Rep;i þ 0:06 Rep;i ÞPri
l0 a;i

The physical system considered is a single droplet
3:5 < Rep;i < 76000 exiting from the nozzle until successful adhesion
for 0:71 < Pri < 380 onto the particle surface or reactor wall or until
1:0 < ðla;i =l0a;i Þ < 3:2 ð14Þ the droplet has been fully evaporated (Fig. 4).

The forces acting on the droplets are gravity,
Because of the relatively small difference between the air buoyancy, and drag.
temperature (Ta,i) and the particle surface temperature
Each droplet remains a unique entity until it is
(Tp,i), the term in Eq. (14), la;i =l0a;i  1. either spray-dried or collected upon the particle
In Eqs. (2), (3), (8), and (9), RD is the drying rate and is or reactor wall surface. So, droplet coalescence is
expressed as mass unit of water evaporated per mass unit not taken into account.
of particles per time unit. The drying rate is related to
During its flight, the droplet retains a spherical
the vapor concentration gradient across the boundary shape, but the droplet size is variable due to water
layer:[24]
a0p;i Ap ðPv;p;i  Pv;a;i Þ
RD;i ¼ ð15Þ
Mp R=MWv ðTa;i þ Tp;i =2Þ
In Eq. (15) the term Pv,p,i is the vapor pressure at the
particle surface and is based on the saturated vapor press-
ure at the particle surface temperature Tp,i. The term
ðTa;i þ Tp;i Þ=2 corresponds to the film temperature[24] and
R=MWv is the specific gas constant for water vapor,
R=MWv ¼ 461.9 J kg 1K 1. If coating solution was col-
lected on the inner reactor wall, the drying rate was calcu-
lated using the modified equation:
a0wall;i Awall;i ðPv;wall;i  Pv;a;i Þ
R1
D;i ¼
  ð16Þ
T þT
R=MWv a;i 2 wall;i

The mass transfer coefficient a0p;i was calculated through an

approximation by means of the dimensionless Sherwood
number, of which the calculation is analogous to Eq. (14):
FIG. 4. The physical system considered in modeling the droplet dynam-
1=2 2=3 2=5 ics along with the jet velocity profiles produced by the pneumatic nozzle as
Shp;i ¼ 2 þ ð0:43Rep;i þ 0:06Rep;i ÞSci ð17Þ described by Schlichting et al.[34]
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1497

evaporation before droplet impact on a bed if the same droplet moves in an infinite fluid in the absence
particle. of solid particles from the fluidized bed. The effective drag

Considering that the diameter of the droplets pro- coefficient, CD, can be estimated based on an empirical
duced in fluidized bed coating is generally lower equation proposed by Mostoufi and Chaouki:[29]
than 40 mm,[6] the temperature inside the droplets
is considered to be uniform (Bidr << 0.1, see CD ¼ embed  CD

ð23Þ
 0:4
Eq (7)). Likewise, the droplet contents are con- 0:33 dd
m ¼ 3:02 Ar0:22
p Re d;t ð24Þ
sidered to be homogeneously distributed. There- dp
fore, effects such as skin or crust formation
As shown in Eq. (24), the droplet effective drag coefficient
during drying are not taken into account.[25]
is a function of the fluidized bed voidage fraction, the
droplet Reynolds number at terminal velocity, the solids’
Droplet Production and Droplet Motion Archimedes number, and the ratio of the droplet diameter
The diameter of the droplets that exit the pneumatic and the bed particle diameter.[29,30] The standard drag
nozzle depends on a number of factors, relating both to 
coefficient CD in Eq. (23) is calculated according to the
the construction parameters of the nozzle and to the correlation given by Turton and Levenspiel:[31]
properties of the spraying liquid and the atomization air,
such as liquid viscosity and surface tension.[26,27]  24 0:413
CD ¼ ð1 þ 0:173Re0:657
d Þþ ð25Þ
Lefebvre[5] proposed the following model for the calcu- Red 1 þ 16300Re1:09
d
lation of the surface-weighted average droplet diameter
In order to calculate the drag force on a droplet, the
(dd,0), given in micrometers, at the exit of a pneumatic or
velocity profile, ~ vat ðh; xÞ, produced by the release of com-
binary nozzle:
pressed air to assist in the atomization of the coating sol-
 0:4  
csol Jsol 0:4 ution has to be known. As stated by various
dd;0 ¼ 0:48 dor 1 þ authors,[2,32,33] the nozzle jet in a fluidized bed shows many
qat DP2at dor Gat
 0:5   similarities to a free axisymmetric jet, of which the gas
2
lsol Jsol velocity profiles are analogous to Schlichting et al.[34] (see
þ 0:15 dor 1þ ð18Þ
qsol csol dor Gat Fig. 4):
Alternatively, the model proposed by Masters[28] may be 1 1
used for pneumatic nozzles: vat;ax ðh; nÞ   ð26Þ
ðhnoz  hÞ 2=3
ð1 þ 0:25 n2 Þ2
pffiffiffiffiffiffiffi  0:45  
585000 3 csol lsol 1000qat Jsol 1:5 n  0:25 n3
dd;0 ¼ pffiffiffiffiffiffiffi þ 597 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi vat;rd ðh; nÞ  ð27Þ
vat;0 qsol csol qsol qsol Gat ð1 þ 0:25 n2 Þ2
ð19Þ
In Eqs. (26) and (27) is n the dimensionless radial coordi-
The droplet exit velocity (vd,0) is assumed to be equal to the nate with
velocity of the released atomization air near the nozzle ori-
fice.[1] This theoretical maximum velocity is thus calculated 2x
n¼ ð28Þ
as: xnoz ðhÞ
sffiffiffiffiffiffiffiffiffiffiffiffiffi The jet radius produced by the pneumatic nozzle, xnoz(h), is
2 DPat defined as the radius where the radial gas velocity,
vd;0 ¼ vat;0 ¼ ð20Þ
qat vat,rd(h,x), equals zero.
Resulting from the second assumption, the balance of
forces exerted upon the droplet can now be written as: Droplet Thermodynamics
During its flight, the droplet exchanges heat and mass
~
F d ¼ Md~
adr ¼ Md~ F bouy þ ~
gþ~ F drag ð21Þ with the surrounding gas phase (air), of which the tempera-
ture and humidity fields are known from the main model.
In Eq. (21) is the third force component, ~ F drag , the drag The thermal energy required for the evaporation and the
force exerted on the droplet. It is calculated as: convective heating of a single droplet equals the heat trans-
2 2 ferred from the gas phase.[35,36] The resulting heat balance
~ p CD qa dd ð~
va þ~
vat ~
vd Þ
F drag ¼ ð22Þ can then be written as:
8
The effective drag coefficient, CD, is generally greater dTd
 Md Cp;d ¼ ad pdd2 ðTa;Kðhd Þ  Td Þ  Qlat RD Md ð29Þ
than the standard drag coefficient, CD , which is the case dt
1498 RONSSE ET AL.

The operator Kðhd Þ in Eq. (29) determines in which con- Using the two-phase model, the bubble voidage fraction,
trol volume, as defined in the main model, the droplet eb, can be obtained from:[22]
resides at height hd. The convective heat transfer coefficient va  vmf
ad was approximated using the dimensionless Nusselt eb ¼ ð36Þ
vb
number, which was calculated according to Eq. (14). The
In Eq. (36) is vb, the bubble rise velocity, and vmf, the mini-
droplet drying rate, RD , was calculated using Eq. (15).
mum fluidization velocity. The minimum fluidization velo-
However, for small spherical geometries, such as droplets
city can be calculated according to the Ergun equation:[39]
in the micron range, the vapor pressure at the droplet sur-
face, Pv,d, in Eq. (15) also depends on the diameter. This  
qa ðqp  qa Þgdp3 1  emf vmf dp qa
vapor pressure correction is given by the Kelvin-Laplace ¼ 150 3
l2a emf la
equation:[37]  
! 1 vmf dp qa 2
þ 1:75 3 ð37Þ
2 Pv;d RTd qw emf la
c ¼ ln 1 ð30Þ
ðdd =2Þ w Pv;d MWw
In Eq. (36), the bubble rise velocity, vb, is calculated
The specific heat of the spraying liquid, Cp,d in Eq. (29), using the correlation proposed by Davidson and
was calculated based on the dry matter concentration in the Harrison:[40]
droplet:
vb ¼ va  vmf þ 0:711ðgdb Þ0:5 ð38Þ
Cp;d ¼ DMd Cp;d þ ð1  DMd ÞCp;c ð31Þ
Finally, the bubble diameter, db, in Eq. (38) is calculated
The change in droplet mass is given by the drying rate: according to the correlation:[41,42]
pffiffiffiffiffiffi
dMd
¼ RD Md ð32Þ db ¼ 0:54g0:2 ðva  vmf Þ0:4 ðh þ 4 Ag Þ0:8 ð39Þ
dt
In Eq. (39) is Ag, the surface area per gas distributor
As the mass of coating material in the droplet is invariable, perforation. When using p porous in stead of perforated
the dry matter content was calculated as: ffiffiffiffiffiffi
gas distributors, the term 4 Ag ¼ 0:03 in Eq. (39). Finally,
DMd;0 Md;0 the bed height was calculated by means of the volume of
DMd ¼ ð33Þ the particle phase and the mean bed voidage, ebed, resulting
Md
from Eqs. (34) to (39).

Droplet Collision Droplet Collection Efficiency

To complete the modeling of the droplet trajectories, During the approach of a droplet toward a particle, dif-
certain operative parameters need to be known, including ferent mechanisms, such as diffusion and inertia, may
the nozzle height, hnoz, with respect to the fluidized bed result in the separation of the droplet trajectory and the
height, hbed; the bed voidage, ebed; and the droplet collec- fluid streamline (Fig. 5), leading to an impact.[1,6] The type
tion efficiency, f, which expresses the probability that a of droplet collision mechanism depends on both the size
liquid drop is successfully collected by a particle, when and the velocity of the impacting droplet and the receiving
their paths meet.[38] particle. Assuming typical droplet size distributions in flui-
dized bed coating, droplet collection caused by inertia is the
Bed Voidage Fraction and Bed Height
In the two-phase model, the mean bed voidage is
given by:[22]
ð1  ebed Þ ¼ ð1  emf Þð1  eb Þ ð34Þ
In Eq. (34) is emf, the bed voidage at minimum fluidization
velocity, while eb is the bubble voidage fraction. The bed
voidage at minimum fluidization velocity, emf, can be calcu-
lated using the Broadhurst and Becker correlation:[39]
!0:029 !0:021
0:7 l2a qa
emf ¼ 0:586 w ð35Þ
qa gðqp  qa Þdp3 qp FIG. 5. The effective cross-sectional area for defining the droplet
impingement efficiency.
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1499

predominant collision mechanism. The ability of a droplet adhesion probability:

to come into contact with a particle is expressed by the
f d ¼ vd 1 d ð43Þ
impingement efficiency vd, which is determined as the ratio
of the effective and the geometrical cross-sectional area:
 2 CALCULATION METHOD
dim
vd ¼ ð40Þ A hybrid solving method was chosen, which combined
dp
a Monte Carlo approach for the particle exchange and
The impingement efficiency can be calculated using the coating collection events with the numerical solving for
following empirical equation:[2,38,43] the remaining ordinary differential equations in the heat
 b and mass balances of the three different phases (gas, solids,
Std and droplets) involved. The advantage of this method over
vd ¼ ð41Þ
Std þ a a complete numerical solving of all differential equations is
that the model is easy to implement and to modify, because
The parameters a and b depend on the particle Reynolds
discretization problems that hinder the direct integration of
number Rep, as shown in Table 1. From Eq. (41) and
the differential equations are not an issue.[44] However, the
Table 1 it can be seen that the impingement efficiency
drawbacks of this method include sample size–related pro-
decreases with decreasing droplet diameter or with decreas-
blems and longer calculation times.
ing gas velocity.
In order to simulate particle exchange, Nsim, particles
Once the droplet has made initial contact with the particle
were chosen to fill the control volumes Si. During each
surface, consecutive interaction between the droplet and the
simulated time step, particles were exchanged with the
particle determines whether the droplet remains adhered to
neighboring control volumes according to the exchange
or bounces off the particle surface. The fraction of droplets
rates ri. Particle exchange was accomplished by random
that remain permanently adhered on the particle surface is
selection of (ri þ ri1)Ni particles in Si and by copying the
expressed by the adhesion probability, 1d , and depends on
selected particles to vacant locations in the control volumes
the droplet impact kinetic energy and the wettability of the
Si1 and Siþ1. In this respect, the described procedure inte-
particle substrate.[2,38] The critical impingement velocity
grates the population (particle) balance equation of each
for droplets impacting on a flat, nonporous, and dry surface
control volume and the solution obtained in this manner
was described by Link,[1,2] droplets are reflected from the
is directly comparable to that obtained from standard inte-
particle surface above this critical velocity:
gration techniques.[44]
4ld ð3 tanðhd=p =2Þ þ tan3 ðhd=p =2ÞÞ2=3 Due to obvious computational limitations, the number
vcrit ¼ ð42Þ of particles used in the simulated population, Nsim, was
dd qd tan2 ðhd=p =2Þ
smaller than the number of particles involved in the real
From Eq. (42) it can be seen that the critical impinge- (physical) system being simulated (Nbed). As the particle-
ment velocity not only depends on droplet characteristics related heat and mass transfers were calculated based on
such as viscosity, size, and density, but also depends on the number of simulated particles, these transfers had to
the static liquid=solid contact angle, hd=p . Contrary to the be multiplied with a scaling factor L to be able to connect
impingement efficiency, the adhesion probability decreases them with gas phase- and droplet phase–related transfer
with increasing droplet size and increasing droplet velocity. terms. In case of a monodisperse particle population, the
The droplet collection efficiency, fd, resulting from scaling factor equals:
successful droplet impingement and droplet adhesion is
given by multiplying the impingement efficiency and Mbed Nbed
L¼ ¼ ð44Þ
Msim Nsim

TABLE 1 As shown in Fig. 6, the particle exchange step preceded

The parameters a0 and b0 for the calculation the solving of the heat and mass balances within each con-
of the droplet impingement efficiency trol volume. These heat and mass balances are ordinary
Rep a0 b0 first-order differential equations and were solved using
the first-order Euler’s method. The particle heat and mass
<1 0.65 3.7 balances, calculated on an individual particle basis, were
10 1.24 1.95 solved prior to the mass and heat balances of both the
40 1.03 2.07 droplet and gas phase. Next, the mass and heat balances
60 0.506 1.84 were integrated over the simulated population and multi-
>>100 0.25 2 plied with the scaling factor from Eq. (44).
1500 RONSSE ET AL.

control volume Si is calculated based on Eq. (6):

DMdr;iþ1 Jdr;iþ1  DMdr;i ðJdr;i
 RD;i Mdr;i  R1 C;i Awall;i Þ  Jsd;i
Mdmd;i ¼ Dtsim ð45Þ
L
Secondly, simulation droplets were sampled using uni-
formly distributed random numbers transformed with the
inverse of the cumulative droplet mass and the cumulative
droplet dry matter content distributions. The droplets’ dry
matter and solvent (water) were then added to randomly
selected particles from the control volume Si. The total
number of droplets sampled in each control volume Si
was calculated based on the amount of dry matter, Mdmd,i,
to be dispersed in Si.

EXPERIMENTAL SETUP
The experimental validation setup is depicted in Fig. 7.
The tapered reactor had a bottom diameter of 0.225 m, a
top diameter of 0.45 m, and a total height of 0.84 m. Fluidi-
zation air was provided by a 2.2-kW high-pressure centri-
fugal fan (Ventomatic CHT166-2T, Merelbeke, Belgium)
of which the air flow was controlled using an invertor.
The air distributor consisted of a Robusta 172 36 wpi
wire mesh (Spörl KG, Sigmaringendorf, Germany). A
pneumatic nozzle (Schlick Model 970-S1, Untersiemau,
Germany) was installed at the tip of a retractable rod,
allowing adjustable nozzle height. The nozzle’s air volu-
metric flow rate V_ at (in Nm3s 1), was obtained through
FIG. 6. Overview of the main model’s simulation procedure. measuring the linear air velocity with a Testo 0.1 m diam-
eter anemometer (Testo, Ternat, Belgium) while blowing
compressed dry air through the nozzle at different atomi-
Prior to calculating the heat and mass balances,
zation pressures (0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, and 4 bar)
the droplet submodel simulation is executed. The air tem-
in a cylindrical pipe with a diameter of 0.1 m and a length
perature and air relative humidity of each control volume
of 0.5 m. Regression analysis resulted in the following
Si are given as input to the droplet submodel and the
correlation between Pat and V_ at :
following submodel calculated data was fed back to
the main model: V_ at ¼ 104 P2at þ 1:3873 103 Pat

The droplet/air heat transfer term adr;i Ni Adr;i and þ 1:7867 103 with R2 ¼ 0:9966 ð46Þ
the droplet/air mass transfer term RD;i Mdr;i .
Micropearl1 glass beads (Sovitec, Fleurus, Belgium)

The droplet mass flow rate Jdr,i and the spray dry-
with d43 ¼ 365 mm were used as the core material. For ther-
ing rate Jsd,i. The rate at which droplet mass
modynamic analysis, distilled water was used as the spray-
adheres onto the particles in control volume
ing liquid. Steady-state spatial air temperature and air
Si ¼ Jdr;iþ1  Jdr;i  Jsd;i  RD;i Mdr;i  R1
C;i Awall;i . relative humidity distributions within the fluidized bed were

The temperature Tdr,i and dry matter content DMdr,i
measured by means of a moving J-shaped probe. Probe
for the droplets in each control volume Si. Due to the
positioning was performed by means of a positioning table
repeated simulation of individual droplets in the sub-
(2 degrees of freedom), computer controlled using optical
model, the resulting droplet property variables (Tdr,i,
rotational encoders for probe positioning control.
DMdr,i and Mdr,i) have their own statistical distri-
A defined quantity of reactor air was drawn through the
bution in each control volume Si.
probe and air temperature was measured by means of a
For the simulation of the coating liquid deposition onto thermistor, while relative humidity was measured by a
the simulated particle population, the following procedure capacitive humidity sensor (Honeywell HIH-3610, Brus-
was used: First, the mass of dry matter (coating polymer) sels, Belgium) at the top end of the probe. The moving
to be dispersed onto the simulated particle population of probe sampled air along the points of a regular grid with
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1501

FIG. 7. Experimental validation setup with the scanning air temperature and humidity probe.

a 0.01 m cell spacing. During each experiment, all process Figure 8a presents the average simulated steady-state
parameters (such as inlet air temperature and flow rate, outlet air temperature (Ta,out) and its standard deviation
liquid flow rate) were kept constant. The scanning was as a function of the number of simulated particles, while
initiated when steady-state operation was reached. The Fig. 8b shows the effect on the simulated steady-state par-
generated spatially distributed data was postprocessed ticle temperature distribution (Tp). The number of simu-
using kriging to generate 2D maps. lated particles strongly influenced the outlet air
temperature, of which the standard deviation reached an
RESULTS AND DISCUSSION asymptotic value of 0.044 C at approximately Nsim ¼
10,000. Considering the particle temperature distribution,
Model Performance no significant influence of Nsim could be found beyond
Error propagation results from the stochastic nature of Nsim ¼ 2500. Therefore, it was concluded that for future
the particle exchange simulation procedure and the droplet simulations, Nsim ¼ 10,000 resulted in sufficiently low stan-
impingement occurrence calculation. Therefore, model sen- dard deviations for the model-predicted output variables.
sitivity studies were performed to investigate the effect of The influence of the number of control volumes on the
the different model design parameters on the statistical simulated steady-state outlet air temperature (Ta,out) is
error of the model predicted variables. The following shown in Fig. 9a, while the effect on the simulated
model design parameters were studied: the number of con- steady-state particle temperature (Tp) is given in Fig. 9b.
trol volumes (n), the number of simulated particles (Np,sim), Increasing the number of control volumes had little influ-
and the number of simulated droplets during each droplet ence on the simulated outlet air temperature. The standard
submodel simulation run (Ndr,sim). The outlet air tempera- deviations on both the simulated outlet air and particle
ture (Ta,out) and particle temperature (Tp) at steady-state temperature increased with increasing number of control
were selected as output variables in this model sensitivity volumes, until a stable value was reached for about 24 to
study. These two variables were chosen because they are 32 control volumes. Based on these results and considering
representative of both the group of variables at control the increasing computational load with increasing n, 24
volume level (Ta,out) and the group of variables at particle control volumes were chosen for all simulations presented
level (Tp). in this study.
1502 RONSSE ET AL.

FIG. 8. (a) The effect of the number of simulated particles (Np,sim) on the steady-state outlet air temperature, Ta,out (-&-) and its standard deviation
(-.-); (b) The effect of the number of simulated particles on the steady-state average particle temperature, Tp,i (-&-) and particle temperature standard
deviation (-.-).

For the remaining model design parameter—the number stationary conditions were reached. The operational and
of modeled droplets during each simulation run (Ndr,sim)— modeling parameters are listed in Table 2.
a series of simulations were performed with the number of In the model Eqs. (1), (2), (5), and (8) is ri the variable
simulated droplets ranging from 100 to 6500. Figures 10a describing the axial particle mixing behavior in the flui-
and 10b present the simulated steady-state temperatures dized bed. The values for the particle exchange rates ri used
of both outlet air (Ta,out) and particles (Tp) as a function in this example calculation were based on measured axial
of the number of simulated droplets (Ndr,sim). In both dispersion coefficients found in the literature. The relation-
graphs it can be seen that the error on the model-predicted ship between the particle exchange rate, ri, and the axial
outlet air temperature and the standard deviation on dispersion coefficient is given by the equation
the particle temperature becomes stable at about
Ndr,sim ¼ 2500 to 3000. Dp ¼ r h2bed ð47Þ
Mostoufi and Chaouki[45] demonstrated that the axial
dispersion coefficients ranged between 3.3 10 3 and
Example Calculation 5.6 10 2 m2s 1 for sand particles with dp ¼ 385 mm and
For a detailed analysis of the model-predicted results, a with superficial gas velocities varying from 0.5 to
water spraying process was simulated from startup until 2.8 ms 1. These operational parameters approximate the

FIG. 9. (a) The effect of the number of control volumes (n) on the steady-state outlet air temperature, Ta,out (-&-) and its standard deviation (-.-);
(b) The effect of the number of control volumes on the steady-state average particle temperature, Tp,i (-&-) and particle temperature standard
deviation (-.-).
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1503

FIG. 10. (a) The effect of the number of simulated droplets (Ndr,sim) on the steady-state outlet air temperature, Ta,out (-&-) and its standard deviation
(-.-); (b) The effect of the number of control volumes on the steady-state average particle temperature, Tp,i (-&-) and particle temperature standard
deviation (-.-).

ones used in this study’s experimental validation and The results of an example calculation using the oper-
example calculation; i.e., glass beads with dp ¼ 365 mm ational and modeling parameters from Table 2 are given
and a superficial gas velocity of 1.19 ms 1. Using Eq. (47) in Figs. 11 and 12. The simulated outlet air temperature
and the results from Mostoufi and Chaouki,[45] the axial (Ta,out) and humidity (Xa,out) during steady-state coating
dispersion coefficient translated into a particle exchange regime, compared with the simulated average particle tem-
rate r ¼ 2.166 Hz. Alternatively, the exchange rate can be perature (Tp,i) and moisture content ðW i Þ, is presented in
calculated based on the average circulation time, for which Fig. 11. To compare air humidity with particle moisture
an equation has been proposed by Rowe[46] for particles in content, the normalized particle moisture content is used:
a bubbling fluidized bed: qp ebed
Wi ¼ ð50Þ
qa;i ð1  ebed Þ
1 0:6ðva  vmf Þð1  ðva  vmf Þðva  vmf Þ=vb Þ
r¼ ¼ ð48Þ
tcn hmf According to the model-predicted results in Fig. 11, con-
stant axial temperature gradients for both the fluidizing air
and the particles were observed, except for the lower
In Eq. (48) is hmf, the bed height at minimum fluidiza- 0.015 m of the fluidized bed. The predicted temperature
tion velocity. It was calculated using Eq. (35). The mini- gradient was primarily determined by the particle exchange
mum fluidization gas velocity, vmf, and bubble rise rate, ri, which is the variable determining the bed’s axial
velocity, vb, were calculated using Eqs. (37) and (38). Using mixing behavior. In the region closest to the air distributor
the conditions summarized in Table 2, the particle (0  h  0:015 m), the air temperature decreased strongly
exchange rate was calculated to be r ¼ 1.78 Hz. as opposed to the average particle temperature. From
For the calculation of the heat transfer rate between Fig. 12, it can be seen that in this region, particles mainly
the fluidized bed and the inner reactor wall, the bubble absorb energy (heat) from the fluidizing air. Once the
frequency, fb, is required. The bubble frequency was esti- heated particles migrate toward the upper bed region,
mated based on Eq. (39) to calculate the bubble diameter where spraying and droplet collection takes place, their
and by taking into account the bubble phase air mass energy is retransferred to the air. The existence of this
flow rate. Assuming the two-phase model for bubbling flui- so-called heat transfer zone, situated close to the air
dized beds,[22] the bubble-phase air mass flow rate, Gb, is distributor and characterized by high air temperatures,
proportional to the superficial gas velocity and was calcu- has been stated by Smith and Nienow,[11] Maronga and
lated as: Wnukowski,[12,13,47] and Heinrich et al.[38]
In the upper half of the bed, the simulated average par-
Gb ¼ Ga;in  ðAbed vmf qa;in Þ ð49Þ ticle temperature is slightly higher than the air temperature
due to the evaporative cooling taking place in the spraying
Based on Eqs. (37) and (49), the bubble frequency was region. The size of the spraying region, which is the volume
found to be around 141 Hz. of the bed in which both droplets and particles coexist,[11]
1504 RONSSE ET AL.

TABLE 2
Model and operational parameters of the simulation
Bed material (glass beads)
Overall mass, kg Mbed 2
Particle diameter, mm dp 365
Density, kgm 3 qbed 2600
Specific heat capacity, Jkg 1 K 1 Cp,p 837
Thermal conductivity, Wm 1 K 1 kp 0.8
Air supply
Mass air flow rate, kg dry air s 1 Ga,in 0.0545
Inlet temperature, C Ta,in 70
Inlet humidity, kg water kg dry air 1 Xa,in 0.00726
Liquid spraying
Spraying rate, kgs 1 Jsol 0.00167
Dry matter content DMsol 0.10
Solution temperature, C Tsol 20
Viscosity coating solution, mPas lsol 1.013
Surface tension coating solution, Nm rsol 0.072
Atomization air mass flow rate, kg dry air s 1 Gat 0.00204
Atomization air temperature, C Tat 20
Atomization relative air humidity uat 0.30
Atomization air mixing constant c 3
Nozzle height, m hnoz 0.126
Nozzle orifice diameter, m dor 0.0008
Other parameters
Reactor wall thickness, m dw 0.002
Reactor wall thermal conductivity, Wm 1 K 1 kw 14.6
Reactor wall emittance for far-infrared radiation e0w 0.16
Reactor wall specific heat, Jkg 1 K 1 Cp,w 500
Reactor wall density, kgm3 qw 8000
Wall mixing constant aw 0.05
Bed voidage at minimum fluidization velocity emf 0.39
Bed voidage ebed 0.82
Bed height, m hbed 0.093
Bubble frequency, Hz fb 141
External temperature, C Te 20
External relative humidity ue 0.5
Model parameters
Control volumes n 24
Number of simulated particles Nsim 20016
Particle exchange rate, Hz r 2.166
Simulation parameters (droplet submodel)

Simulation droplets=run Nsim 2500
Simulation time step Dtsim adaptive, range:
1 ms!0.1 ms
Simulation parameters (main model)
Simulated time, s tsim 1500
Simulation time step, s Dtsim 0.001
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1505

FIG. 11. Simulated air temperature, Ta,i (-&-) and air absolute FIG. 13. The simulated fraction of non-collided droplets (-&-) as a
humidity, Xa,i (-.-) of the fluidized bed during steady-state spraying (coat- function of height, compared to the linear droplet concentration model.[43]
ing) regime, combined with the average particle temperature, Tp,i (-&-)
and the average, normalized particle moisture content, W i (- -).

that nearly all of the collected coating liquid was evapo-

rated inside the spraying region. Normally, the drying
can be calculated from Fig. 13, in which the fraction of free region should extend deeper into the bed as opposed to
(non-collided) droplets, Mdr ðhÞ=Mdr ðhnoz Þ, is plotted the spraying zone,[12,13] but due to use of the properties
against their height above the air distributor. According of water in modeling, the droplet size according to Eq.
to Löffler,[43] the concentration of the droplet phase in a (18)—resulting in very small droplets (dd ¼ 15; 2 mm using
spray decreases linearly with the axial distance to the spraying liquid parameters in Table 2)—in combination
nozzle, or with the low spraying rate (Jsol ¼ 3.63 gmin 1) and the
Mdr ðhÞ 3 ð1  ebed Þfd h moderate fluidization air flow rate (Gat ¼ 0.0545 kgs 1),
 ð51Þ water evaporation was modeled to be very fast
Mdr ðhnoz Þ 2 ebed dp hnoz
(a0p ¼ 313:5 ms1 during steady-state spraying regime).
From Fig. 13 it can be seen that a quasi-linear relation- The model was also capable of predicting the dynamic
ship exists between the fraction of free (or noncollided) coating mass distribution (Y p) and the coating efficiency
droplets and the axial distance to the nozzle. By comparing (dc). The coating efficiency is defined as the fraction of
the heat transfer rates due to evaporation in Fig. 12 with coating mass collected on the surface of the core particles,
the size of the spraying region in Fig. 13, it was concluded compared to the total amount of coating mass injected into
the fluidized bed.[48–51] Or, assuming no attrition and using
Eq. (9), the coating efficiency is calculated as:

PNbed Pn Pn
i¼1Yp;i i¼1 Jsd;i þ i¼1 R1
C;i Awall;i DMdr;i
dc ¼ ¼1
Jsol DMsol Dtprocess Jsol DMsol
ð52Þ

In this example calculation, the predicted coating

efficiency was around 95%. However, by modifying certain
operational variables, the coating efficiency could be
altered significantly, which is demonstrated into more
detail in the next section. The model-predicted coating
mass distribution is given in Fig. 14. During steady-state
spraying regime, the relationship between the predicted
average coating mass and the process time is linear, as
FIG. 12. Simulated heat exchange rates per control volume during demonstrated in the literature.[1,7,52,53] Initially, the simu-
steady-state spraying (coating) regime: particle=gas phase heat exchange
rate (-&-), evaporative heat exchange at particle surface (-.-), droplet=gas
lated coating mass distribution was asymmetric but
gas phase heat exchange rate (-&-), and heat loss through the reactor wall evolved to a normal distribution, of which the standard
toward the environment, (- -). deviation is proportional to the square root of the process
1506 RONSSE ET AL.

average model-predicted particle temperature and

particle moisture content remain constant, unlike
their distributions. Particle moisture content distri-
bution is particularly useful, as it correlates with
the agglomeration tendency of the particles.[8]

Spray drying losses. Considering the reactor
geometry in Table 2, coating losses due to reactor
wall contact, R1 C , were negligible compared to
spray-drying losses. Consequently, the relative
spray-drying loss equals 1  dc , according to
Eq. (52).

The average droplet penetration depth into the
fluidized bed,
hd , and its standard deviation,
r(hd). The standard deviation of the coating
growth rate, rðd_ p Þ. The average growth rate
remains constant when modifying spraying vari-
ables, unlike its distribution. The narrower the
growth rate distribution, the more uniform the
coating layer,[55] and, consequently, the higher
the quality of the produced coated particulate
solids.
FIG. 14. Predicted evolution of the coating mass distribution (Y), In the sensitivity analysis, the following effects were
expressed as kg coating per kg core material.
found:

time, which is in agreement with the conclusions drawn by
Atomisation air pressure, Pat. The effects of the
Maronga and Wnukowski[47] and Nakamura et al.[54] atomization air pressure on droplet collection
mechanisms and bed thermodynamics are com-
plex. Firstly, increasing the atomization air press-
Sensitivity Analysis
ure will also increase the volumetric air flow rate
The effects of varying the values of several variables,
of cooler and drier atomization air; see Eq. (46).
more specifically variables related to the spraying nozzle
Consequently, the air temperature in the upper
and coating liquid, on the coating process was studied
part of the bed decreases, which in turn slows
using the proposed model. To describe the sensitivity of
down the droplet evaporation rate. The contri-
the model, each variable was modified individually and
bution of the atomization air to the increase of
the translation factor, 1=K, was calculated using the fol-
the bed’s air temperature gradient is also the main
lowing equation:
reason of the rather high translation factors in
1 @X x Fig. 15a when using the standard deviation of
¼  ð53Þ
K X @x the particle temperature, r(Tp), as the output
variable X. Secondly, as shown by Lefebvre[5]
The translation factor describes the relative change of an
and Juslin et al.,[27] increasing the air pressure also
output parameter X as a result of a relative change of the
decreases the droplet diameter and increases the
parameter x. A translation factor higher then 1 indicates
initial droplet velocity. The relationship between
an amplifying effect of the parameter variation; a value
the droplet size and droplet evaporation rate is
lower than 1 implies an attenuating effect of the parameter
nearly linear;[56] furthermore, smaller droplets tend
variation.
to decelerate faster and will therefore be more sus-
In the sensitivity analysis, the following output variables
ceptible to complete evaporation. Finally, the
(X) were considered (Fig. 15):
droplet collection probability, fd, depends on the

The standard deviation of both the particle tem- droplet size. Whereas smaller droplets have—
perature dsitribution, r(Tp), and the particle according to Eq. (41)—decreased impingement
moisture content distribution, r(Wp), during efficiency (vd), their adhesion probability (1d ) is
steady-state spraying regime. Changing the spray- increased. As a result, the average droplet path
ing variables does not alter the overall thermodyn- length is reduced with increasing atomization air
amic operation point of the bed; consequently, the pressures (Figs. 15d and 15e). In Fig. 16, the
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1507

FIG. 15. Model sensitivity, represented as the translation factors ( , Dx > 0; , Dx < 0) using the following output variables: the standard deviation
of the particle temperature (Fig. 15a), standard deviation on particle moisture content (Fig. 15b), spray drying losses (Fig. 15c), average droplet
penetration depth (Fig. 15d), standard deviation of droplet penetration depth (Fig. 15e), and standard deviation of particle (or coating layer) growth
rate (Fig. 15f).

modeled droplet penetration distribution is given losses or highest coating efficiency. This effect
for different atomization air pressures. could be reproduced using the presented model.
Due to these complex and counteracting phenom- Figure 17 shows the model-predicted spray-drying
ena involving atomization air pressure, the effect losses, compared to the reference conditions of
of varying the atomization air pressure on the Pat ¼ 2.5 bar, as a function of atomization air
spray-drying losses, and hence the coating pressure. According to the model-predicted
efficiency, is somewhat less pronounced (Fig. results, a pressure of Pat ¼ 3 bar yielded minimal
15c). According to the coating experiments of spray-drying losses.
Dewettinck and Huyghebaert[15] and Saleh et
Nozzle height, hnoz. Figure 18 shows the results for
al.,[7] an optimal atomization air pressure could the effects of nozzle position (hnoz) on the droplet
be found, which resulted in minimal spray-drying penetration depth distribution in a fluidized bed
1508 RONSSE ET AL.

FIG. 18. Model-predicted droplet penetration depths (in m above

FIG. 16. Model-predicted droplet penetration depths (in m above air air distributor) with varying nozzle positions: median penetration depth
distributor) with varying atomization air pressures: median penetration (-&-); 10th (- -) and 90th (-&-) percentile of droplet penetration depth.
depth (-&-); 10th (- -) and 90th (-&-) percentile of droplet penetration
depth.
seen that spray-drying losses continue to increase
significantly, even beyond hnoz ¼ 0.13 m.
with a bed height hbed ¼ 0.093 m. From a certain
Spraying rate, Jsol. The spraying rate is the princi-
nozzle height (hnoz > 0.13 m) there is little influ- pal variable influencing the bed’s temperature and
ence on the droplet penetration depth. By increas- humidity. Increasing the spraying rate yields
ing the distance between the nozzle and the higher temperature and humidity gradients in the
fluidized bed, the volume of spraying region—the fluidized bed;[7] hence, the large translation factors
region where collision between particles and drop- when using the standard deviation of the particle
let occurs—is reduced and, consequently, the par- temperature (Fig. 15a) and particle moisture con-
ticle temperature and moisture content have wider tent (Fig. 15b). Increasing the spraying rate also
distributions, as shown in Figs. 15a and 15b. The reduced the droplet evaporation rate because of
pronounced translation factors for droplet depth the higher humidity and lower temperature in
distribution in Figs. 15d and 15e results from the the bed, resulting in lower spray-drying losses
fact that the reference condition used hnoz ¼ (Fig. 15c) and slightly deeper droplet penetration
0.10 m, a region were strong correlation exists depths (Figs. 15d and 15e).
between nozzle height and droplet penetration
Coating solution viscosity, lsol. According to
depth. By comparing the results from Fig. 18 with Eq. (18), a linear relationship exists between the
the spray-drying losses shown in Fig. 19, it can be viscosity of the coating solution and the surface-
weighted average diameter of the produced

FIG. 17. Model-predicted spray-drying losses as a function of atomi- FIG. 19. Model-predicted spray-drying losses as a function of nozzle
zation air pressure, expressed relatively to the reference case of Pat ¼ 2.5 bar. position, expressed relatively to the reference case of hnoz ¼ 0.093 m.
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1509

droplets. Consequently, increasing the viscosity

significantly reduces the spray-drying losses as
shown in Fig. 15c. Increased droplet viscosity also
results in larger adhesion probabilities, 1d ;[57,58]
however, according to Eq. (42) this effect is neutra-
lized by the decreasing droplet diameter. As a
result, the coating solution viscosity has no signifi-
cant effect on the average droplet penetration
depth (Fig. 15d) and has therefore little effect on
the size of the spraying region, which in turn
explains the low sensitivity when using the distri-
bution width of both the particle moisture content,
Wp, and particle (or coating layer) growth rate, d_ p ,
as output variable X.
FIG. 20. Model-predicted spray-drying losses as a function of fluidiza-

Coating solution surface tension, csol. Similar to the tion air flow rate, V_ in , expressed relatively to the reference case of
coating solution, increasing the surface tension of V_ in ¼ 0.015 m3s 1.
the coating solution results in the production of
larger droplets, which in turn will give rise to an
increased volume of the spraying region, wider spraying region (Fig. 15d) and less uniformity in
particle temperature and moisture content distri- particle temperature (Fig. 15a) are predicted.
butions (Figs. 15a and 15b), and less spray-drying
Atomization air temperature, Tat. The atomization
losses (Fig. 15c). Compared to the coating solution air temperature contributes to the air temperature
viscosity, the effect of the coating solution surface and humidity gradient in the fluidized bed. Conse-
tension on the size of the spraying volume is more quently, altering the temperature of the atomi-
pronounced, as can be seen in Figs. 15d and 15e. zation air has an, albeit less pronounced, effect
This observation could be explained by the fact on the spray-drying losses (Fig. 15c) and the size
that surface tension, as opposed to viscosity, does of the spraying region (Figs. 15d and 15e).
not alter the droplet adhesion probability, 1d , as
Fluidization air flow rate, Ga,in. Adjusting the flui-
demonstrated in Eq. (42). The results dealing dization air flow rate changes the bed height, hbed,
with the influence of csol are in agreement with and the particle exchange rates, ri. Therefore, the
the experimental observations of both Panda position of the nozzle was adjusted in such a way
et al.[57,58] and Hemati et al.,[7] who reported that that the nozzle tip coincided with the bed height
coating solution surface tension is supposed to in each simulation and the particle exchange rate
have a higher influence on droplet deposition was recalculated using Eq. (48). Increasing the
and particle growth compared to the viscosity of air flow rate, Ga,in, significantly increases the evap-
the coating solution. orative capacity of the bed and thus increases the

Coating solution initial temperature, Tsol. In overall spray drying loss, as shown in Figs. 15c
Fig. 15c, it can be seen that the temperature at and 20. Similar conclusions regarding coating
which the coating solution is injected into the flui- efficiency were drawn by Saleh et al.[10] from their
dized bed has a large impact on the amount of fluidized bed coating experiments with varying
spray-drying losses. A first explanation lies in the fluidization air flow rates. It is also to be expected
expression of the input variable, x, in Eq. (53) that the volume of the spraying region is reduced
being the solution temperature (Tsol) in Kelvin when using a higher evaporation rate (related to
rather then degrees Celsius and thus the term Gin). However, increasing the fluidization air
@x=x in Eq. (53) is significantly smaller. Secondly, flow rate also increases the bed height and, thus,
assuming the droplet drying process is adiabatic, the bed voidage fraction, ebed, which reduces the
an additional drying loss of 0.17 wt% is obtained volumetric droplet collection probability. The net
per Kelvin temperature difference between the result of both effects is complex, as can be seen
initial droplet temperature and the adiabatic dry- in the large asymmetry for both (positive and
ing temperature. This additional drying loss due negative parameter variation) translation factors
to the initial droplet temperature explains the large when using the standard deviation on droplet
effect on the overall spray drying losses in Fig. 15c penetration depth, r(hd), as output variable X in
and, as a result, a slightly smaller volume of the Fig. 15e.
1510 RONSSE ET AL.

Experimental Validation
Figures 21 and 22 illustrate the measured steady-state
air temperature and relative humidity distributions in a
fluidized bed of 2 kg glass beads, with inlet air temperature
Ta,in ¼ 50 C and a spraying rate of 3.63 g=min 1. Con-
sidering the temperature profiles, no radial temperature
gradients were observed in the fluidized bed, which is in
agreement with the results obtained by Heinrich et al.[38]
In the humidity profiles, a region with significantly higher
air humidity is found below the nozzle. This so-called
spraying region is the zone of the bed where both particles
and droplets coexist and where droplet=particle mass trans-
fer (adhesion) takes place. The existence of this spraying
region has been reported by Smith and Nienow,[11]
Maronga and Wnukowski,[12,13,47] Heinrich et al.,[38] and
Hemati et al.[7]
Considering the fact that the overall model is one-dimen-
sional, the measured spatial air temperature and humidity
distributions were integrated into one-dimensional axial
profiles. Figures 23 and 24 show the comparison between
the measured and model-predicted steady-state axial tem-
perature profiles using two different inlet air temperatures
(Ta,in ¼ 50 C and Ta,in ¼ 70 C) and using three different
spraying rates (Jsol ¼ 3.63 g=min 1, 5.24 g=min 1, and FIG. 22. Measured steady-state spatial air relative humidity profile of a
6.85 g=min 1). From Fig. 23, it can be seen that there is a water-sprayed fluidized bed of glass beads (Ta,in ¼ 50 C, Jsol ¼
rather good agreement between the modeled and measured 3.63 gmin 1).
axial temperature profiles, although the model predicts
slightly higher air temperatures near the air distribution
plate. When using a higher inlet air temperature (Fig. 24),
the model predicted more uniform temperature gradients
compared to the measured axial temperature gradients.

CONCLUSIONS
A model has been presented enabling the calculation of
the dynamic behavior of a top-spray fluidized bed coater.
The model combines the one-dimensional discretized

FIG. 21. Measured steady-state spatial air temperature profile of a FIG. 23. Measured versus simulated steady-state axial air temperature
water-sprayed fluidized bed of glass beads (Ta,in ¼ 50 C, Jsol ¼ profile of a water-sprayed fluidized bed of glass beads with Ta,in ¼ 50 C
3.63 gmin 1). and with varying spraying rates.
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1511

to control the formation of out-of-specification waste pro-

ducts and to reduce operational costs, it is necessary to
incorporate additional phenomena, such as particle
agglomeration and attrition, into the model. The discrete
nature of the model and its flexible architecture render
the implementation of these side effects easy to tackle.

NOMENCLATURE
A Surface (m2)
Ar Archimedes number [¼ d 3 qf ðqs  qf Þg=l2f ]
a Acceleration (ms 2)
0
a Impingement efficiency parameter
Bi Biot number
FIG. 24. Measured versus simulated steady-state axial air temperature b0 Impingement efficiency parameter
profile of a water-sprayed fluidized bed of glass beads with Ta,in ¼ 70 C
and with varying spraying rates.
CD Drag coefficient, corrected for fluidized beds

CD Drag coefficient
Cp Specific heat (Jkg 1K 1)
c Atomization air mixing constant
representation of the fluidized bed with a spraying submo- D Dispersion or diffusion coefficient (m2s 1)
del that serves to calculate the individual droplet trajec- DM Dry matter content (kg dry matter kg 1
tories along with the heat and mass transfers between the solution)
droplet and the surrounding air. The model was shown d Diameter (m)
to be capable of predicting the dynamic axial temperature, F Force (N)
humidity, and coating distributions during the top-spray G Air mass flow rate (kg dry air s 1)
fluidized bed coating process. Furthermore, the model g Gravimetric constant, 9.81 ms 2
may also be used to asses the extent to which unwanted h Height (m)
spray drying of the coating solution occurs. I Cycles
A sensitivity analysis was carried out to study the J Mass flow rate (kgs 1)
influence of spraying-related process variables on coating 1=K Translation factor (sensitivity analysis)
process characteristics, such as spray-drying losses, coating L Scaling factor
growth, and droplet penetration depth. Although the M Mass (kg)
major process variables, being used to set the coating pro- MW Molecular weight (kgmol 1)
cess’ thermodynamic operation point such as inlet air flow m Correction factor for the effective drag
rate, spraying rate or inlet air temperature, have the highest coefficient
significance in determining the coating characteristics, less N Number of particles
obvious and often overlooked parameters—such as coating Nu Nusselt number [¼ a d=kf 
solution and nozzle properties—have shown to be as N Number of droplets
important. Within the range of conditions examined and n Number of control volumes
considering spraying-related variables and=or parameters, P Pressure (Pa)
the coating solution injection temperature and viscosity Pr Prandtl number [¼ lf Cp;f =kf ]
were shown to have the largest impact on the spray drying Pv Vapor pressure (Pa)
losses due to premature droplet evaporation. When con- P1 v Vapor pressure at flat surface (Pa)
sidering the spraying volume, its size was mainly determ- Q Heat (J)
ined by the solution surface tension, followed by the R Universal gas constant, 8.314 Jkg 1K 1)
nozzle height and atomization pressure. Nozzle height RC Droplet collection rate (kg solution s 1kg
and coating solution surface tension were also shown to core 1)
1
be the main parameters in determining the coating layer RC Droplet collection rate at reactor wall
growth rate variability and—along with the atomization air (kg solution m 2s 1)
pressure—the temperature uniformity of the fluidized bed. RD Drying rate (kg water s 1 kg core 1)
The dynamic nature of the model allows its use in simu- RD 
Droplet drying rate (kg water s 1 kg droplet 1)
1
lating responses of the system to a change in the working RD Droplet drying rate at reactor wall (kg water
conditions and therefore it could be applied to assess m 2s 1)
new and advanced control strategies. However, in order Re Reynolds number [¼ dvf qf =lf ]
1512 RONSSE ET AL.

ri Rate of particle exchange from Si toward ax Axial

Siþ1, fraction of population s 1 b Bubble
S Control volume bed Fluidized bed
Sc Schmidt number [¼ lf =qf D] bouy Buoyancy
Sh Sherwood number [¼ a0 d=D] c Coating
St Stokes number [¼ qd va dd2 ð9la dp Þ1 ] cn Circulation
T Temperature (K) cond Conduction
t Time (s) conv Convection
V Volume (m3) d Individual droplet
v Velocity (ms 1) dmd Dispersed dry matter
W Particle moisture content (kg water kg core 1) dr Droplet phase
W Normalized particle moisture content (kg water drag Drag
kg dry air 1) e Exterior
X Absolute air humidity (kg water kg dry air 1) f Fluidum
x Radial distance from the jet axis (m) g Gas (air) distributor
Y Collected coating mass (kg dry matter im Impingement
kg core 1) in Inlet
lat Latent
loss Heat loss
Greek Symbols
mf Minimum fluidization
a Convective heat transfer coefficient noz Nozzle
(Wm 2K 1)
or Nozzle orifice
a0 Mass transfer coefficient (ms 1)
out Outlet
d Coating efficiency (overall)
p Particle
c Surface tension (Nm)
rad Radiation
Dd Thickness (m)
rd Radial
Dt,Dt Time step
real Real, referring to the system being
e Porosity
simulated
e0 Emissivity s Solid (droplet and=or particle)
f Droplet collection efficiency
sd Spray dried
g Fraction of heat loss originating from
sim Simulated
particles compared to total heat loss
sol Coating solution
h Contact angle (static)
t Terminal
K Control volume membership
v Vapor
k Thermal conductivity (Wm 1K 1)
w Water
l Viscosity (Pas)
l0 Viscosity at particle=droplet REFERENCES
surface (Pas)
1. Link, K.C.; Schlünder, E.-U. Fluidized bed spray granulation.
n Dimensionless radial coordinate Investigation of the coating process on a single sphere. Chemical
q Density (kgm 3) Engineering and Processing 1997, 36 (6), 443–457.
1 Adhesion probability 2. Zank, J.; Kind, M.; Schlünder, E.-U. Particle growth and droplet
r Standard deviation deposition in fluidised bed granulation. Powder Technology 2001,
120 (1–2), 76–81.
r Stefan-Boltzmann constant, ¼ 5.669 10 8 W
3. Desai, K.G.H.; Park, H.J. Recent developments in microencapsula-
m 2K 4 tion of food ingredients. Drying Technology 2005, 23 (7), 1361–1394.
v Impingement efficiency 4. Dewettinck, K.; Huyghebaert, A. Fluidized bed coating in food
w Sphericity technology. Trends in Food Science and Technology 1999, 10 (4–5),
U Heat transfer rate (Js 1) 163–168.
5. Lefebvre, A.H. Atomization and Sprays; Taylor & Francis: New York,
u Relative humidity
1988.
X Output variable (sensitivity analysis) 6. Guignon, B.; Duquenoy, A.; Dumoulin, E.D. Fluid bed encapsulation
x Input variable (sensitivity analysis) of particles: Principles and practice. Drying Technology 2002, 20 (2),
419–447.
7. Hemati, M.; Cherif, R.; Saleh, K.; Pont, V. Fluidized bed coating and
Subscripts
granulation: Influence of process-related variables and physico-
a Fluidization air chemical properties on the growth kinetics. Powder Technology
at Atomization air 2003, 130 (1–3), 18–34.
 NUMERICAL SPRAY MODEL OF FLUIDIZED BED COATING 1513

8. Smith, P.G.; Nienow, A.W. Particle growth mechanisms in fluidized- 30. Wang, X.; Zhu, C. Concentric evaporating spray jets in dilute
bed granulation. I. The effects of process variables. Chemical Engin- gas-solids pipe flows. Powder Technology 2003, 129 (1–3), 59–71.
eering Science 1983, 38 (8), 1223–1231. 31. Turton, R.; Levenspiel, O. A short note on the drag correlation for
9. Becher, R.D.; Schlünder, E.-U. Fluidized bed granulation—The spheres. Powder Technology 1986, 47 (1), 83–86.
importance of a drying zone for the particle growth mechanism. 32. Donadono, S.; Mareca, A.; Massimilla, L. Gas injection in shallow
Chemical Engineering and Processing 1998, 37 (1), 1–6. beds of fluidized, coarse solids. Ingenieria Chimica Italiana 1980,
10. Saleh, K.; Cherif, R.; Hemati, M. An experimental study of fluidized- 16, 1–10.
bed coating: Influence of operating conditions on growth rate and 33. Becher, R.D.; Schlünder, E.-U. Fluidized bed granulation: Gas flow,
mechanism. Advanced Powder Technology 1999, 10 (3), 255–277. particle motion and moisture distribution. Chemical Engineering
11. Smith, P.G.; Nienow, A.W. On atomizing a liquid into a gas-fluidized and Processing 1997, 36 (4), 261–269.
bed. Chemical Engineering Science 1982, 37 (6), 950–954. 34. Schlichting, H.; Gersten, K.; Krause, E.; Oertel, H.; Mayes, C. Bound-
12. Maronga, S.J.; Wnukowski, P. Establishing temperature and humidity ary Layer Theory, 8th Ed; Springer-Verlag: Berlin, 2004.
profiles in fluidized bed coating. Powder Technology 1997, 94 (2), 35. Chen, Z.; Wu, W.; Agarwal, P.K. Steam-drying of coal. Part 1.
181–185. Modeling the behavior of a single particle. Fuel 2000, 79 (8), 961–973.
13. Maronga, S.J.; Wnukowski, P. The use of humidity and temperature 36. Gupta, A.; Rao, D.S. Model for the performance of a fluid catalytic
profiles in optimizing the size of fluidized bed in a coating process. cracking (FCC) riser reactor: Effect of feed atomization. Chemical
Chemical Engineering and Processing 1998, 37 (5), 423–432. Engineering Science 2001, 56 (15), 4489–4503.
14. Jones, D.M. Factors to consider in fluid-bed processing. Pharmaceu- 37. Scherer, G.W. Adsorption in aerogel networks. Journal of
tical Technology 1985, 9, 50–62. Non-Crystalline Solids 1998, 225 (1–3), 192–199.
15. Dewettinck, K.; Huyghebaert, A. Top-spray fluidized bed coating: 38. Heinrich, S.; Blumschein, J.; Henneberg, M.; Ihlow, M.; Peglow, M.;
Effect of process variables on coating efficiency. Food Science and Mörl, L. Study of dynamic multi-dimensional temperature and con-
Technology 1998, 31 (6), 568–575. centration distributions in liquid-sprayed fluidized beds. Chemical
16. Teunou, E.; Poncelet, D. Batch and continuous fluid bed coating— Engineering Science 2003, 58 (23–24), 5135–5160.
Review and state of the art. Journal of Food Engineering 2002, 53 39. Rhodes, M.J. Introduction to Particle Technology; John Wiley and
(4), 325–340. Sons: Chichester, 1998.
17. Dewettinck, K.; De Visscher, A.; Deroo, L.; Huyghebaert, A. Model- 40. Cryer, S.A. Modeling agglomeration processes in fluid-bed
ing the steady-state thermodynamic operation point of top-spray granulation. AIChE Journal 1999, 45 (10), 2069–2078.
fluidized bed processing. Journal of Food Engineering 1999, 39 (2), 41. Darton, R.C.; LaNauze, D.C.; Davidson, J.F.; Harrison, D. Bubble
131–143. growth due to coalescence in fluidized beds. Transactions of the
18. Schaafsma, S.H.; Kossen, N.W.F.; Mos, M.T.; Blauw, L.; Hoffmann, Institutions of Chemical Engineers 1977, 55, 274–280.
A.C. Effects and control of humidity and particle mixing in fluid-bed 42. Patil, D.J.; Van Sint Annaland, M.; Kuipers, J.A.M. Critical compari-
granulation. AIChE Journal 1999, 45 (6), 1202–1210. son of hydrodynamic models for gas–solid fluidized beds—Part II:
19. Straatsma, J.; Van Houwelingen, G.; Steenbergen, A.E.; De Jong, P. Freely bubbling gas-solid fluidized beds. Chemical Engineering
Spray drying of food products: 1. Simulation model. Food Engineer- Science 2005, 60 (1), 73–84.
ing 1999, 42 (2), 67–72. 43. Löffler, F. Staubabscheiden; Thieme Verlag: Stuttgart, 1988.
20. Collier, A.P.; Hayhurst, A.N.; Richardson, J.L.; Scott, S.A. The heat 44. Lin, Y.L.; Lee, K.; Matsoukas, T. Solution of the population balance
transfer coefficient between a particle and a bed (packed or fluidised) equation using constant-number Monte Carlo. Chemical Engineering
of much larger particles. Chemical Engineering Science 2004, 59 (21), Science 2002, 57 (12), 2241–2252.
4613–4620. 45. Mostoufi, N.; Chaouki, J. Local solid mixing in gas-solid fluidized
21. Janna, W.S. Engineering Heat Transfer, 2nd Ed; CRC Press: Boca beds. Powder Technology 2001, 114 (1–3), 23–31.
Raton, FL, 2000. 46. Rowe, P.N. Estimation of solids circulation rate in a bubbling
22. Kunii, D.; Levenspiel, O. Fluidisation Engineering, 2nd Ed; fluidized bed. Chemical Engineering Science 1973, 28 (3),
Butterworth-Heinemann: Stoneham, MA, 1991. 979–980.
23. Sparrow, E.M.; Abraham, J.P.; Tong, J.C.K. Archival correlations 47. Maronga, S.J.; Wnukowski, P. Modelling of the three-domain
for average heat transfer coefficients for non-circular and circular fluidized-bed particulate coating process. Chemical Engineering
cylinders and for spheres in cross-flow. International Journal of Heat Science 1997, 52 (17), 2915–2925.
and Mass Transfer 2004, 47 (24), 5285–5296. 48. Taranto, O.P.; Rocha, S.C.S.; Raghavan, G.S.V. Convective heat
24. Campbell, G.S. An Introduction to Environmental Biophysics; transfer during coating of tablets in two-dimensional spouted-bed.
Springer-Verlag: New York, 1977. Drying Technology 1997, 15 (6–8), 1909–1918.
25. Farid, M. A new approach to modelling of single droplet drying. 49. Donida, M.W.; Rocha, S.C.S. Coating of urea with an aqueous
Chemical Engineering Science 2003, 58 (13), 2985–2993. polymeric suspension in a two-dimensional spouted bed. Drying
26. Schæfer, T.; Wørts, O. Control of fluidized bed granulation. III. Technology 2002, 20 (3), 685–704.
Effects of inlet air temperature and liquid flow rate on granule size 50. Vieira, M.G.A.; Donida, M.W.; Rocha, S.C.S. Adhesion of an aque-
and size distribution. Control of moisture content of granules in the ous polymeric suspension to inert particles in a spouted bed. Drying
drying phase. Archives of Pharmaceutical and Chemical Science, 5th Technology 2004, 22 (5), 1069–1085.
Edition. 1978, 6, 1–13. 51. Oliveira, H.V.A.; Peixoto, M.P.G.; Freitas, L.A.P. Study on the
27. Juslin, L.; Antikainen, O.; Merkku, P.; Yliruusi, J. Droplet size efficiency of hard gelatin capsules coating in a spouted bed. Drying
measurement: I. Effect of three independent variables on droplet size Technology 2005, 23 (9–11), 2039–2053.
distribution and spray angle from a pneumatic nozzle. International 52. Guignon, B.; Regalado, E.; Duquenoy, A.; Dumoulin, E. Helping to
Journal of Pharmaceutics 1995, 123 (2), 247–256. choose operating parameters for a coating fluid bed process. Powder
28. Masters, K. Spray Drying Handbook; John Wiley and Sons: Technology 2003, 130 (1–3), 193–198.
New York, 1991. 53. Saleh, K.; Steinmetz, D.; Hemati, M. Experimental study and model-
29. Mostoufi, N.; Chaouki, J. Prediction of effective drag coefficient ing of fluidized bed coating and agglomeration. Powder Technology
in fluidized beds. Chemical Engineering Science 1999, 54 (7), 851–858. 2003, 130 (1–3), 116–123.
1514 RONSSE ET AL.

54. Nakamura, H.; Abe, E.; Yamada, N. Coating mass distributions of seed kinetics in a fluidized bed. Chemical Engineering and Processing
particles in a tumbling fluidized bed coater Part II. A Monte Carlo 2004, 43 (6), 693–699.
simulation of particle coating. Powder Technology 1998, 99 (2), 140–146. 57. Panda, R.C.; Zank, J.; Martin, H. Experimental investigation of
55. Larsen, C.C.; Sonnergaard, J.M.; Bertelsen, P.; Holm. P. A new process droplet deposition on a single particle. Chemical Engineering Journal
control strategy for aqueous film coating of pellets in fluidised bed. 2001, 83 (1), 1–5.
European Journal of Pharmaceutical Sciences 2003, 20, 273–283. 58. Panda, R.C.; Zank, J.; Martin, H. Modeling the droplet deposition
56. Leclère, K.; Briens, C.; Gauthier, T.; Bayle, J.; Guigon, P.; behavior on a single particle in fluidized bed spray granulation
Bergougnou, M. Experimental measurement of droplet vaporization process. Powder Technology 2001, 115 (1), 51–57.