chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

A one-dimensional plug-flow model of a counter-current

spray drying tower

Muzammil Ali a , Tariq Mahmud a,∗ , Peter John Heggs a , Mojtaba Ghadiri a ,
Dusan Djurdjevic b , Hossein Ahmadian b , Luis Martin de Juan b , Carlos Amador b ,
Andrew Bayly b
a Institute of Particle Science and Engineering, The University of Leeds, Engineering Building, Leeds LS2 9JT, UK
b P&G Technical Centres Ltd, Whitley Road, Longbenton, Newcastle Upon Tyne NE12 9BZ, UK

a b s t r a c t

A one-dimensional numerical model for a detergent slurry drying process in a counter-current spray drying tower
is developed for the prediction of the gas and droplet/particle temperature profiles within the tower. The model
accommodates droplets/particles over a range of sizes. A semi-empirical slurry droplet drying model is integrated
with a counter-current tower simulation based on mass, energy and particulate phase momentum balances in order
to calculate the drying rate and the particle residence time within the tower. The coupled first order ordinary dif-
ferential equations for the two phases are solved numerically using the iterative shooting method in an algorithm
developed within MATLAB. The predictions of the numerical model are compared with industrial pilot plant data.
The results are found to vary significantly with the specified size distribution of the droplets. Despite the simplic-
ity of the model in ignoring the coalescence, agglomeration, wall deposition and re-entrainment, the model gives
reasonable agreement with the experimental data.
© 2013 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Spray drying; Slurry droplet drying; One dimensional spray tower model; Counter-current spray tower
model; Heat and mass transfer; Plug flow model

1. Introduction The gas loses most of the heat at the top in evaporating the
moisture from the droplets. Counter-current spray drying tow-
Spray drying is one of the most common methods for the ers are used for thermally stable products like detergents and
manufacture of a wide variety of particulate products in the ceramics as the hot gas comes in contact with dry particles just
chemical, food, household care products, pharmaceutical and before the particles exit the tower. These spray drying towers
ceramics industries. Spray drying involves the evaporation of are more thermally efficient because of better utilisation of
the solvent from a solution or slurry containing solids using a heat due to counter-current contact between the two phases
hot gas to transform droplets into dry particles. Spray drying is and a higher temperature difference exists between the two
carried out in a tower, in which the solution or slurry is sprayed phases at most of the tower height. This study deals with
into small droplets using nozzles. The droplets come in con- modelling of a counter-current spray drying tower. Fig. 1 is
tact with a co-current or counter-current hot gas flow and a schematic diagram of a counter-current spray drying tower.
simultaneously exchange heat, mass and momentum, result- The moisture content, bulk density, particle size dis-
ing in the solid particle formation in a single unit (Masters, tribution and temperature are important properties and
1972). Co-current spray drying towers are suitable for drying parameters of the dried powder products. The modelling
of heat sensitive products as the hot gas comes in contact of a spray drying tower enables prediction of these prop-
with the droplets at the top where the droplets have maxi- erties of the dried powder, and hence, the optimisation of
mum moisture content and are at the wet bulb temperature. the drying operation. This requires modelling the drying of

∗
Corresponding author. Tel.: +44 113 34 3243.
E-mail address: [email protected] (T. Mahmud).
Received 18 March 2013; Received in revised form 9 July 2013; Accepted 9 August 2013
0263-8762/$ – see front matter © 2013 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cherd.2013.08.010
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 827

Nomenclature
drop of the droplet
Ap surface area of droplet/particle (m2 ) gas of the drying gas
CD drag coefficient init initial value
Cv,s surface vapour concentration (kg/m3 ) ins of the insulation
Cv,∞ vapour concentration in the bulk (kg/m3 ) l of the liquid
Cl,s moisture concentration at the surface (kg/m3) s of the solid
Cl,i initial moisture concentration (kg/m3 ) p of the droplet/particle
cp specific heat (J/kg K) slurry of the slurry
d diameter (m) solid of the solid
dm size constant (m) term terminal
DAW diffusion coefficient of water into the air (m2 /s) vap of the vapours
DWS diffusion coefficient of water in the slurry w of the wall
(m2 /s) water of water
F a buoyancy force (N) z in the z direction
F d drag force (N)
g acceleration due to gravity (m2 /s) Dimensionless constants
˛d
hfg latent heat of vapourisation of moisture in Nu Nu = gasp , Nusselt number
droplet (J/kg) Pr Pr =
gas cp,gas
, Prandtl number
gas
H specific humidity gas dp |vp −vgas |
kc mass transfer coefficient (m/s) Re Re = gas , Reynolds number
gas
K partition coefficient Sc Sc = gas DAW , Schmidt number
M mass (kg) kc dp
Sh Sh = DAW , Sherwood number
Ṁ mass flow (kg/s)
N molecular weight (g/mol)
n number per unit time (1/s)
and the particles. The modelling of spray drying towers is
p pressure (Pa)
very challenging, because of the complexity associated with
Q̇ measured slurry volumetric flow rate (m3 /s)
the integration of the drying of the droplets with the inter-
Rg gas constant (J/mol K)
acting hydrodynamic and heat/mass and momentum transfer
r droplet radius (m)
between the two phases. This is further exaggerated due to
rc air core radius (m)
interactions between the droplets/particles (resulting in coa-
ro nozzle orifice radius (m)
lescence/agglomeration/breakage) and interaction between
t time (s)
tsd surface drying time (s)
T temperature (K) Solution/slurry inlet
T̃ mass weighted average temperature (K) Exhaust air
us distribution parameter
U overall heat transfer coefficient (W/m2 K)
v velocity (m/s)
ṽ velocity at the inlet of a control volume (m/s)
wl moisture fraction
w̃l mass weighted average moisture fraction
y probability density function
Yd cumulative frequency oversize
z vertical distance (m)
Z tower height (m)

Greek letters
˛ heat transfer coefficient (W/m2 K)
 density (kg/m3 ) Hot air inlet Hot air inlet
 thermal conductivity (W/m K)
 viscosity (kg/m/s)
ı thickness (m)
 spray cone angle

Subscripts
amb ambient
boil boiling
D of the tower

Dried Particles
droplets/particles inside the tower by considering interactions
of the transport processes between the droplets/particles and Fig. 1 – Schematic diagram of a counter-current spray
the hot gas. Spray drying involves complex flow of the gas drying tower.
828 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

the droplets/particles and wall (resulting in wall deposition).

To capture these processes in a rigorous way, an advanced
modelling effort combining CFD with DEM and population
balance would be required to account for inter-particle colli-
sion, coalescence, agglomeration, deposition on the wall and
re-entrainment of the particles from the wall. However, this
would be highly computationally intensive. It is highly desir-
able to have a simpler model when the role of droplet drying
model, input process variables, residence time and temper-
ature profiles needs to be assessed in a reasonably short
time.
Many researchers have studied the drying behaviour of
a single droplet experimentally (Charlesworth and Marshall,
1960; Cheong et al., 1986; Furuta et al., 1983; Hassan and
Mumford, 1993; Hecht and King, 2000a; Lin and Gentry, 2003;
Nesic and Vodnik, 1991; Ranz and Marshall, 1952; Sano and
Keey, 1982; Sunkel and King, 1993). The data obtained from
the experimental observations have enabled the development
of mathematical models for single droplet drying. Ranz and
Marshall (1952) were the first one to investigate the evapo-
ration of water droplets containing dissolved and suspended
solids. For the droplets containing dissolved solids, they con-
cluded that before the formation of a solid structure, the
solvent evaporates at a constant rate and the solution is satu-
rated throughout the droplet. The evaporation rate is less than
that for a pure solvent. For the droplets containing suspended
solids, the presence of solid particles does not lower the
vapour pressure significantly; hence, the initial evaporation
rate corresponds to that of pure solvent. After the formation
of a solid structure, the falling rate period begins and the
temperature rises continually. Since then, many researchers
cited above have carried out further single droplet drying
experiments for various solutions and slurry droplets. These Fig. 2 – Mechanisms of droplet drying models.
experiments revealed that the drying behaviour depended
strongly on the morphological development, which in turn
depended primarily on the composition of the droplet.
The drying of droplets containing dissolved or suspended
solids has commonly been divided into two distinct stages
2. Droplet with a pliable crust and a bubble (Handscomb et
(Masters, 1972). In the first stage, the liquid evaporates from
al., 2009; Hecht and King, 2000b; Sano and Keey, 1982).
the surface at a fairly constant rate. The size of the droplet
reduces due to moisture evaporation. The second stage begins
when a solid crust covers the surface of the droplet. The drying
rate in this stage becomes entirely dependent on the internal The modelling approaches based on these two mecha-
diffusion of moisture to the surface. The moisture content at nisms differ in the second stage of the drying process. In the
which the transformation from the droplet to solid occurs is type 1 model, it is assumed that a rigid porous solid crust is
termed as critical moisture content. Hecht and King (2000b) formed and the outer diameter of the particle remains con-
introduced a third stage which occurs when the particle tem- stant. The evaporation only occurs at the interface between
perature reaches the boiling point of slurry. The drying rate in the crust and the wet core. The vapour then diffuses through
this stage is controlled by the rate of heat transfer to the par- the porous solid crust to its outer surface. The thickness of
ticle. It should be noted that in spray drying particles with a the crust increases as solid is deposited on the internal crust
wide range of morphologies can be produced which have been core interface due to evaporation of moisture (Fig. 2(a)). The
described by various phenomenological models (see review in wet core continues to shrink until all the moisture is evapo-
Charlesworth and Marshall, 1960; Walton and Mumford, 1999). rated resulting in a dry porous particle as shown in Fig. 3(a). In
Two mechanisms of droplet drying process have been used for the type 2 model (Fig. 2(b)), after the formation of pliable solid
the development of mathematical models in previous stud- crust, the moisture diffuses to the outer surface of the parti-
ies, herein referred to as type 1 and type 2 models, as detailed cle from where it evaporates throughout the drying process.
below: An increase in temperature of the particle causes vapouri-
sation of solvent within the wet core. It is assumed that a
single centrally located saturated vapour bubble expands due
1. Droplet with a dry solid crust (Audu and Jeffreys, 1975; to vapourisation of the solvent, which increases the internal
Cheong et al., 1986; Dalmaz et al., 2007; Elperin and pressure of the particle. This causes the particle to inflate.
Krasovitov, 1995; Farid, 2003; Handscomb et al., 2009; Kadja An arbitrary maximum particle size is specified in order to
and Bergeles, 2003; Mezhericher et al., 2007, 2008; Nesic and limit the expansion of the particle. This approach explains the
Vodnik, 1991). formation of dry hollow particles (Fig. 3(b)).
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 829

1 Particle
Particle
0.9 Droplet
0.8 Droplet
0.7
0.6
0.5

Yd
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2
Diameter (mm)

Fig. 4 – Size distribution on a cumulative mass basis.

distribution. The model results were presented for a co-

current spray drying tower. It was concluded that in the model
with a single droplet size representing the mean diameter, the
particles were dried out in a much shorter time. Montazer-
Rahmati and Ghafele-Bashi (2007) proposed a mathematical
model for a counter-current spray drying of detergent slurry
droplets. The model assumes plug-flow of the particles and
the gas phase. The droplet/particle size distribution is rep-
resented through a fixed mean diameter. The drying of the
particles takes place using type 1 modelling approach. The
heat loss from the column is neglected. A unique feature of
their model is entrainment of finer particles by the exit gas
stream. The fines form 6% to 8% of the dried product leaving
the tower. The exchange of particles from the downward mov-
ing particle stream to the upward moving particle stream takes
place when particles of a certain fraction of the distribution
in the downward moving particle stream reach the terminal
velocity.
In this paper, a mathematical model for a counter-current
spray drying tower is developed using the plug-flow approach.
The model simulates drying of detergent slurry droplets
of varying sizes using a semi-empirical single droplet dry-
ing model developed in-house by P&G (Hecht, 2012), based
on the model proposed by Hecht and King (2000b). The
plug flow model considers axial variation of temperature
Fig. 3 – Scanning electron microscope images of spray and humidity of the hot gas. The density and viscosity of
dried particles. (a) Porous ceramic particle (source: Effting et the hot gas are allowed to vary as a function of temper-
al. (2010)), (b) Hollow detergent particle. ature. The model is validated against an industrial pilot
plant spray drying tower and can be used to predict and to
The single droplet drying model needs to be coupled with optimise the operation of counter-current spray drying tow-
the various transfer mechanisms occurring within the spray ers.
tower: momentum, heat and mass, so that the spray dried
powder properties can be predicted and the operation of the 2. Size distribution
tower optimised. Parti and Palancz (1974) developed a math-
ematical model for particle drying in both co-current and The spray drying process starts from a distribution of droplets
counter-current spray drying towers that considers mono- resulting from the atomisation of feed using an atomiser. The
sized particles, uniformly dispersed over the cross-section of estimation of initial size distribution may be important in pre-
the spray drying tower. The droplet/particle and air velocities dicting the overall performance of the spray drying tower. In
are one-dimensional and parallel to the axis of the dryer, it this study, the measured size distribution of the slurry droplets
is also called plug-flow approach, in which complete mixing as well as the measured size distribution of the dried powder
along the cross-section and no back mixing is assumed. The is used as the initial size distribution by considering two cases.
temperature and humidity of air are allowed to vary axially. The size distribution of the droplets is measured using laser
The drying model considers shrinking of the droplet due to diffraction. The size distribution of the dried powder is mea-
evaporation of moisture. The diameter is kept constant once sured by sieving. The cumulative size distribution curve of the
it reaches the critical moisture content. The heat loss from sprayed droplets in the simulation is represented by using a
the column is neglected. Topar (1980) extended the model Rosin–Rammler fit (Rosin and Rammler, 1933) given by Eq. (1).
proposed by Parti and Palancz (1974) by adding droplet size Fig. 4 is a plot of the measured size distribution on cumulative
830 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

First Stage Second Stage

A B C D

Surface Evaporation Crust Formation Bubble Nucleation

Initial Slurry Droplet

G F E

Final Particle Equilibrium Inflated Particle

Moisture Content
Fourth Stage Third Stage

Fig. 5 – Stages of drying of a slurry droplet.

mass basis represented by discrete dots and the continuous reaches equilibrium. In the fourth stage, the particle temper-
lines represent the respective curves fitted using Eq. (1). ature approaches the gas temperature and the equilibrium
moisture in the particle changes accordingly.
 
Yd = exp −(dp /dm )us (1)

3.2. Drying model assumptions

3. Droplet/particle drying model
The following assumptions are applicable to the single
The spray drying of droplets of detergent slurry consisting droplet/particle drying model used in this study:
of an aqueous solution containing insoluble solid particles
is addressed. The droplet/particle drying model is applied to
each particle size. Major components of the detergent slurry 1. There are no temperature/concentration gradients within
include a surfactant, a polymer, a binding agent dissolved the droplet/particle. Since the droplets/particles are very
in water and a softening agent as solid particles. The exact small (ranging from 50 ␮m to 2300 ␮m), the variation of
composition of the slurry cannot be provided because of con- temperature within the droplet can be neglected (Biot num-
fidentiality; however it is very similar to the detergent slurry ber is small <0.1).
studied by Griffith et al. (2008) and Handscomb et al. (2009) in 2. Internal circulation inside the slurry droplet is neglected,
which the slurry had a moisture percentage of 29% w/w. The because the presence of solid particles inside the small
model based on Hecht and King (2000b) has been selected for droplets hampers internal circulation.
describing the drying rate of the detergent slurry in a counter- 3. The droplets and the resulting particles remain spherical
current spray drying tower as it explains the formation of throughout the tower. However, the shape of the particles
hollow particles (see Fig. 3(b)), which is observed in the dried may undergo changes due to morphological development
detergent particles. during drying, agglomeration, particle–wall interaction and
breakage of the particles and are not accommodated at this
3.1. Droplet/particle drying mechanism stage.
4. It is assumed that drying continues to take place in the
The drying process illustrated in Fig. 5 consists of four stages. third stage until all the moisture is evaporated from the
In the first stage (A–C), the initial heating/cooling up of particle, since modelling particle drying in the fourth stage
droplets to the wet bulb temperature takes place as the mois- requires a temperature dependent water isotherm for the
ture evaporates from the surface of the droplet. The size of detergent powder, which is not available. Therefore, in the
droplet also decreases due to the evaporation of water in fourth stage, only sensible heat transfer takes place from
this stage. The second stage (C–D) begins when the water in the gas to the particle.
the droplet is insufficient to maintain a saturated condition 5. To account for particle inflation in the third stage due to
at the droplet surface, thus causing a solid crust to form at internal vapourisation of moisture. The size of the parti-
the surface. The particle temperature begins to rise rapidly cle is increased to the initial diameter of the droplet. This
in this stage. The third stage (D–F) starts when the particle results in a change in particle density. The choice of the
temperature is equal to the boiling point of the slurry. The maximum size of the inflated particle is based on the mea-
particle inflates in this stage due to internal vapourisation sured density of dried powder.
of moisture in a saturated vapour bubble. Drying continues 6. The density and specific heat of the slurry and the diffusiv-
after maximum inflation of the particle until the moisture ity of vapours into the bulk remains constant.
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 831

3.3. Governing equations where erfc is the complementary error function and K is the
partition coefficient, given by
The change in temperature of the droplet/particle as a func-
tion of vertical distance can be calculated by the following Cv ∼ Hgas
K= = (10)
energy balance: Cs wl,initial slurry

dTp Eq. (9) is simplified by introducing the following dimension-

Mp cp,drop ṽp = ˛p Ap (Tgas − Tp )
  dz
    less parameter:
Heat input to the droplet/particle by convection
Heat absorbed by the droplet/particle

dMl t
+ hfg (2) B = K kc (11)
dt
   DWS
Heat consumed in vapourizing the moisture

Eq. (9) now becomes

The heat transfer coefficient, ˛p , in Eq. (2) is calculated from
the Ranz and Marshall (1952) correlation:
(Cl,s − Cl,i )  
= 1 − exp B2 erfc(B) (12)
(Cl,∞ − Cl,i )
Nu = 2.0 + 0.6Re0.5 Pr1/3 (3)

The surface drying time (tsd ) is taken as the time for Cl,s
The moisture content in the droplet is calculated using the
to reduce to a value corresponding to 90% of the equilibrium
equation proposed by Hecht and King (2000b):
moisture content. Thus, Eq. (12) becomes
2
ṽp
dwl
=
(1 − wl ) dMl
(4)
 
dz Msolid dt exp B2 erfc(B) = 0.1 (13)

The change in droplet radius due to evaporation of liquid The solution of Eq. (13) results in
in the first stage of drying (see Fig. 5) is given by

drp dMl /dt B = 5.5 (14)

ṽp = (5)
dz 4l rp 2
From Eqs. (11) and (14), tsd is given by
In the first stage of drying, the evaporation of moisture from
the surface of the droplet depends upon the surface vapour 5.5
2
concentration and the vapour concentration in the bulk. The tsd = DWS (15)
K kc
initial slurry droplet drying rate is given by

Once the surface is dry (step C in Fig. 5), the drying rate
dMl
= −4rp 2 kc (Cv,s − Cv,∞ ) (6) becomes dependent on internal diffusion of moisture to the
dt
surface. Hecht (2012) developed an algebraic equation for this
The surface vapour concentration (Cv,s ) in Eq. (6) is cal- step by fitting results from a full numerical model for droplet
culated by assuming the partial pressure of vapours at the drying previously developed by Hecht and King (2000b). The
surface to be in equilibrium with the liquid phase, it is given drying rate for this stage is given by
by
  B 
psat,vap Nl dMl t − tsd
Cv,s = (7) = −106 rp exp −A −C (16)
1000 × Rg Tp dt 106 rp

The saturated vapour concentration is calculated using

where A, B and C are constants for the exponential curve fit
Antoine’s equation.
with values of 18.9, 0.2 and 17.7, respectively.
kc in Eq. (6) is the mass transfer coefficient, and is calculated
Puffing (particle inflation) starts at the beginning of stage 3,
from the Ranz and Marshall (1952) correlation as follows:
when the particle temperature is equal to the boiling point of
the slurry (step D in Fig. 5). To take into account the inflation of
Sh = 2.0 + 0.6Re0.5 Sc1/3 (8) particles in this stage, the diameter of the particle is changed
to the initial droplet diameter. During the puffing stage, the
The transformation of the first stage of drying to the second drying process is controlled by the external heat transfer from
stage takes place when a layer of solids covers the surface of the gas to the particle. As the particle dries, the boiling tem-
the droplet. The time for the surface to become dry is approxi- perature of the slurry increases. The drying rate is obtained by
mated by solving the diffusion equation in planar coordinates a simple energy balance on a particle, with the boiling tem-
system for diffusion in a semi-infinite slab. The analytical perature of the slurry represented as a function of moisture
solution is given by Crank (1975) and for the surface concen- content:
tration it becomes
 
2
(Cl,s − Cl,i ) Kkc

2 Kk

p
˛p 4
d
(Tgas − Tp )
c 2
= 1 − exp DWS t erfc DWS t dMl
(Cl,∞ − Cl,i ) DWS DWS
dt
=−   dTboil (17)
hfg − ws cp,s + wl cp,l ws dw
(9) l
832 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

The relationship between the boiling point of the detergent

dTgas 
J
nj
slurry and moisture content was determined experimentally Ṁgas cp,gas = ˛j Apj (Tpj − Tgas ) ×
dz ṽp,j
and is given by    j=1
Heat transfer from the air to the droplets   
276.25

Total heat input to the droplets by convection

Tboil = exp − 6.6 + 373.15 J

23.68 + 100wl
(18)  dM nj
l
+ cp,vap (Tpj − Tgas ) × + 2ri U(Tamb − Tgas ) (19)
dt j ṽp,j   
j=1
4. Spray drying tower model    Heat loss to the environment

Sensible heat to the evaporated vapours

The model of the spray drying tower illustrated in Fig. 1 is

where j is the identifier for each discrete size. J is the total
taken to be a vertical cylinder with a uniform cross-sectional
number of discrete sizes. The number of droplets or particles
area. In the actual tower, the hot gas enters the bottom
(nj ) for each discrete size in a control volume per unit time is
enlarged region of the tower by a number of inlets. The tower is
given by
operated at slightly below atmospheric pressure which causes
entrainment of air from the bottom exit of the tower. In the Ṁslurry,j
nj = (20)
model, an enthalpy balance is used to obtain the inlet hot gas Mpj
temperature (adiabatic mixing of hot gas and cold entrained
air streams) and the gas velocity is also a combination of For the calculation of heat loss to the environment, a con-
the two streams. This hot gas flows counter-currently to the stant thickness of column wall and insulation is assumed. The
droplet/particles. The mass flow, humidity, density and tem- overall heat transfer coefficient (U) in Eq. (19) is calculated
perature of gas are allowed to vary in the axial direction along the column height and is given by
by dividing the tower height into a number of equal incre-  r +ıw   ri +ıw +ıins 
1 1 ri ln i r ri ln ri
ments because the resulting differential equations need to = + i
+
ri +ıw
+ (21)
be discretised and solved using a finite difference method. U ˛D w ins ˛amb (ri + ıw + ıins )
In each increment, the heat and mass transfer between the
two phases and the particulate phase momentum transfer are The Nusselt number used for calculation of inside film coef-
solved. ficient (˛D ) is given by (Kreith, 1973)

NuD = 0.023Re0.8
D Pr
0.33
(22)
4.1. Assumptions incorporated in the spray drying
tower model
Eq. (22) is applicable to fully developed flow in a circular
duct. The change in mass flow of the gas is given by the fol-
1. The flow of hot gas in a spray drying tower involves
lowing equation:
complex three-dimensional swirling flow. However, for
simplicity, the radial and angular velocities of the hot gas
 −nj dM

J
and droplets/particles are assumed to be zero. dṀgas l
= (23)
dz ṽp,j dt
2. Hot gas is assumed to follow the ideal gas law. The    j=1
j

spray drying tower operates at atmospheric pressure and Change in mass flow of gas   
medium temperature (200–400 ◦ C). Change in mass of droplets/particles

3. Heat transfer by radiation is neglected as it only becomes

The droplet/particle velocity for each particle size is calcu-
significant at very high temperatures.
lated from the equation of motion:
4. The droplets/particles are uniformly dispersed over the
cross section of the tower with no interaction between
dvp
them. In actual spray drying towers, interaction between Mp ṽp = Mp g
   dz 
the droplets/particles takes place, which may result in droplet/particle weight
Change in droplet/particle momentum
change in size of the droplets/particles. The modelling of
particle–particle and particle–wall interactions inside the + F + F (24)

a

d
spray drying tower is not included due to the complexities
Bouyancy force Drag force
involved.
5. The minimum velocity of the particles is limited to the ter-
minal falling velocity of the particles. Although particles The buoyancy force (F a ) is given by
of diameter up to 200 ␮m get entrained with the gas and  
gas i
exit from the top, but this amount is only 3% by mass of F a = −Mp g (25)
the dried powder collected from the bottom. Hence the p
entrainment of particles is not considered. In the actual
spray drying tower, the particles move close to the wall The drag force (F d ) in Eq. (24) is given by
where the gas velocity is almost zero. This leads to smaller  
1
particles exiting from the bottom which would otherwise F d = − gas rp2 CD vp − vgas  (vp − vgas ) (26)
2
get entrained in the gas stream.
The Reynolds number based on the relative velocity is given
4.2. Governing equations by
 
gas dp vp − vgas 
An energy balance on the gas phase results in the following Re = (27)
equation for the variation of the gas temperature: gas
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 833

The drag coefficient for the droplets is calculated using cor- all the droplet sizes and is calculated using the following
relations proposed by Williams (1976), which was measured equation:
for burning oil droplets:
Q̇
For Re < 80, CD = 27Re−0.84 (28) vp =    (37)
 ro2 − rc2 cos 2

For Re > 80, CD = 0.271Re0.217 (29)

The above equation requires the radius of the air core (rc )
and spray cone injection angle (). The injection angle is taken
The drag coefficient for the particles is calculated using
to be 40◦ based on the vendor provided data and the radius of
correlations proposed by Morsi and Alexander (1972) which
the nozzle ro is 1.38 × 10−3 m. The air core radius is taken from
is applicable to smooth spherical particles. It is given by
the data reported by Nelson and Stevens (1961).
a2 a3
CD = a1 + + 2 (30)
Re Re 4.4. Boundary conditions
where a1 , a2 and a3 are constants for several ranges of parti-
At z = 0 (at the top of the tower corresponding to droplet inlet),
cle Reynolds number. The minimum velocity of the particle
the following boundary conditions are applicable:
is limited to the particle terminal velocity to avoid negative
velocity values for the particles and is given by
Tp = Tp,0 , vp = vp,0 , wl = wl,0 , dp = dp,0 ,

4gdp (p − air ) Ṁgas = Ṁgas,0 , Tgas = Tgas,0 , and vgas = vgas,0 (38)
vp,term = (31)
3CD air

The density of the hot gas as a function of temperature is

At z = Z (at the bottom of the tower corresponding to the
calculated by the ideal gas law:
gas inlet), the following boundary conditions apply:
pNgas
gas = (32)
1000 × Rg Tgas Tgas = Tgas,Z and vgas = vgas,Z (39)

The viscosity of gas is considered as a function of gas tem-

perature and is calculated using the following relationship
5. Solution methodology
obtained from air viscosity data in Perry and Green (1997):
The spray tower is divided into a number of equal sized incre-
gas = 10−5 (0.0036Tgas − 6 × 10−7 Tgas 2 + 1.8626) (33) ments of length z to solve the differential equations using a
finite difference method. The differential Eqs. (2), (4), (5), and
The mass weighted average particle temperature at the (24) are discretised using a forward difference approximation;
outlet is calculated using the following equation: whereas Eqs. (19) and (23) are discretised using a backward
difference approximation. In Eq. (2), the mass, drying rate,

J heat transfer coefficient, specific heat and velocity of each
Mp z,j Tp z,j particle size are considered to be constant within the incre-
j=1 ment. The counter-current particle-gas flow arrangement and
T̃p = (34) unknown exit conditions of the gas (at z = 0) and particles (at

J

Mp z,j z = Z), necessitate an iterative technique to obtain the numer-

j=1 ical solution. The starting point of the calculation is at the
top of the tower (z = 0) where the slurry is sprayed (see Fig. 6).
The mass weighted average particle moisture content at The initial values of the outlet gas temperature and mass flow
the outlet is calculated using the following equation: are required in order to solve the discretised equations for
the first iteration. These values are estimated based on the

J overall energy balance assuming that the particles outlet tem-
Mp z,j wl z,j peratures are equal to the inlet gas temperature and moisture
j=1 fraction of the particles is zero. The calculated gas tempera-
w̃l = (35)

J ture at the bottom of the tower (z = Z) is compared with the
Mp z,j known value of inlet gas temperature. The initial estimated
j=1 value of the outlet gas temperature is then adjusted, based on
the difference between the calculated value and the known
The temperature of the gas stream going into the tower is value of inlet gas temperature. The outlet gas moisture con-
calculated using the following equation: tent is also adjusted, based on the moisture content of the
dried particles. The calculation is repeated until required val-
Ṁgas,hot cp,gas,hot Tgas,hot + Ṁgas,cold cp,gas,cold Tgas,cold ues of tolerances of 0.5 K between the calculated and known
Tgas = (36)
Ṁgas,hot cp,gas,hot + Ṁgas,cold cp,gas,cold gas inlet temperatures as well as a 1% difference between the
outlet mass flow of the gas for two successive iterations is
4.3. Droplets initial velocity obtained. The solution methodology is implemented in the
computer software package MATLAB. Fig. 7 is the logic flow
The slurry is atomised using a hollow-cone pressure noz- diagram of the algorithm for the numerical solution of the full
zle atomiser. The initial velocity is assumed to be same for model.
834 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

Table 1 – Input operating conditions.

Droplet properties

Slurry inlet temperature 365.5 K

Slurry mass flux 0.17 kg/m2 s
Specific heat of dried particle 1500 J/kg K
Specific heat of solvent 4180 J/kg K
Specific heat of vapours 1900 J/kg K
Density of slurry 1566 kg/m3
Latent heat of vapourisation 2.26 × 106 J/kg
Diffusion coefficient of water in 3.0 × 10−11 m2 /s
slurry
Diffusion coefficient of water 2.6 × 10−5 m2 /s
vapour into gas

Gas properties
Hot gas temperature 559.5 K
Hot gas mass flux 0.76 kg/m2 s
Gas pressure 101,325 Pa
Gas thermal conductivity 0.03 W/m K
Entrained air mass flux 0.038 kg/m2 s
Specific heat 1006 J/kg K
Ambient temperature 281 K

Column wall
Metal wall thickness 0.006 m
Metal wall thermal conductivity 18.8 W/m K
Insulation thickness 0.105 m
Insulation thermal conductivity 0.04 W/m K

content and gas temperature, as well as the inlet gas tem-

perature are listed for total number of increments of 3580 and
7160. The sensitivity of the results on the number of grid size
is very small. The grid size selected for all further simulations
is 3580. The simulation run time for 3580 grids is about 1 min
on a desktop PC.
Fig. 6 – Control volume inside the spray tower.
Fig. 8 is a plot of the convergence of the difference between
the predicted and known inlet gas temperatures (these are
6. Simulation results calculated from the measured inlet and the entrained air tem-
peratures) against the iteration number for the simulation
The overall model described in the previous sections has been with 3580 grids. The solution reaches the required tolerance
used to simulate the spray drying of slurry for the generation limit (0.5 K) in seven iterations.
of detergent powders in an industrial counter-current pilot
plant tower with a height to diameter ratio greater than 3. 6.2. Solution dependency on cut sizes
Two cases are simulated to evaluate the effect of specified
size distribution on the simulation results. In the first case The size distribution of the droplets/particles in Fig. 4 is repre-
(Case 1), the experimentally measured size distribution of the sented by a number of discrete sizes. Increase in the number
droplets is used. In the second case (Case 2), the final dried of discrete sizes used to represent the size distribution should
powder size distribution is used as the initial size distribu- result in a more accurate representation of how the size dis-
tion. tribution affects the predicted results of the simulations, but
The data listed in Table 1 are from an experimental run there will be a corresponding increase in the run times per
with a slurry mass flux of 0.17 kg/m2 s entering the tower at simulation. Table 3 lists simulated values of the average par-
365.5 K. The hot gas mass flux is 0.76 kg/m2 s at a tempera- ticle outlet temperature and moisture content, and the gas
ture of 559.5 K and the entrained air mass flux is 0.038 kg/m2 s outlet temperature for decreasing the cut size from 50 down
at a temperature of 281 K. The measured outlet parameters to 12.5 ␮m for the droplet size distribution. The magnitude of
were: outlet gas temperature of 382.2 K, average outlet parti- the cut size influences the simulation results. An increase in
cle temperature and moisture content of 361.7 K and 2.49% the average moisture content and outlet gas temperature can
wt/wt, respectively. be seen, however the influence of size cut on particle aver-
age outlet temperature is negligible. The predictions using
6.1. Solution dependency on the number of increments a cut size of 25 ␮m are very close to the predictions using
(z) 12.5 ␮m. Hence a size cut of 25 ␮m is sufficient for this inves-
tigation.
The solution was investigated for numerical accuracy by The simulation provides the outlet gas temperature as
increasing the total number of increments (z). The particle well as the outlet temperature and moisture content of each
size distribution was represented using 39 discrete sizes. In particle size cut. The mass weighted average outlet powder
Table 2, the variations of the calculated values of the out- temperature and moisture content are evaluated from these
let mass weighted average particle temperature, moisture results. A difference in overall enthalpy balance error is 0.6%
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 835

Start with estimated value of

.
Tgas,out and Mgas,out

Calculate droplet/particle velocity and

Reynolds number

j=j+1

z=z+dz
Calculate drying rate, droplet temperature,
moisture content and new diameter

j<J Yes

No
. . Calculate new
Calculate gas temperature
Mg,k+1=Mg,k (Tgas,z)
value of Tgas,out

No
z>Z

Yes

No (Tgas,calc-Tgas,Z)
<=Tol

Yes

No . . .
(Mgas,k+1-Mgas,k)/Mgas,k x 100 <= 0.1

Yes
End

Fig. 7 – Logic flow diagram of the solution algorithm.

Table 2 – Solution dependency on the number of increments.

S. no. Total number No. of Particle average outlet Particle average Outlet gas
of increments iterations temperature (K) moisture % (w/w) temperature (K)

1 3580 7 525.31 1.08 378.28

2 7160 7 525.32 1.08 378.30

and 0.3% for Cases 1 and 2, respectively, which is based on where E is the enthalpy of a stream and the subscripts 1,
these predicted outlet values and the inlet conditions and 2, 3, 4 and 5 refer to the streams of gas inlet, the slurry
given by the following expression: inlet, the gas outlet, the particles outlet and heat loss,
respectively. Additionally, profiles along the tower are also
provided for the residence times, temperatures, moisture
100[(E1 + E2 ) − (E3 + E4 + E5 )]/(E1 + E2 ) (40) contents, drying rates and velocities for each particle size cut.

Table 3 – Solution dependency on size cut.

S. no. Number of Size cut Particle average Particle average Gas outlet Simulation
discrete (␮m) outlet temperature moisture content % temperature (K) time (s)
sizes (K) (w/w)

1 20 50 525.57 1.07 375.41 39

2 39 25 525.31 1.08 378.28 76
3 77 12.5 525.18 1.09 379.82 156
 836 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

160 Moisture Fraction (wl/wl,0)

0 0.05 0.1 0.15 0.2 0.25 0.3
140 0
50 µm
120 0.1 100 µm

Dimensionless Height (z/Z)

0.2 200 µm
100
Difference (K)

0.3 400 µm
80 0.4 800 µm
60 0.5 1000 µm
0.6
40
0.7
20 0.8

0 0.9
1 2 3 4 5 6 7 8 1
Iteration
Fig. 10 – Moisture profiles of droplets/particles.
Fig. 8 – Plot of the difference in predicted and gas inlet
temperatures vs. Iteration.
as the particles lose the moisture; hence a sharp rise in the
Also profiles of the gas temperature, mass flow and heat flux particle temperature can be seen as the moisture evaporates
are given. from the particles. The larger particles (greater than 400 ␮m)
Case 1: In Case 1, the measured droplet size distribution is exit at lower temperatures, because the residence times of
specified as the initial size distribution. Fig. 9 is a plot of tem- these particles is shorter than the smaller particles and the
perature profiles of the droplets/particles of selected sizes in specific surface area is also smaller. The hot gas tempera-
microns and the hot gas moving counter-current to the parti- ture is highest at the bottom of the tower and decreases as
cles. The initial temperature of the slurry is greater than the the gas flows towards the top due to heat exchange with the
wet bulb temperature (calculated to be 322 K at the exhaust droplets/particles.
gas conditions); therefore, the temperature of all the droplet Fig. 10 is a plot of dimensionless moisture profiles (wl /wl,0 )
sizes start to fall as the droplets move downwards. However, of droplets/particles of selected sizes. It can be seen that the
no droplet attains a constant wet bulb temperature. After a smaller particles sizes lose the moisture quicker compared
certain height, the temperature of particles starts to rise. This to larger particle sizes. This is due to greater residence time
is due to the formation of a crust at the surface and the trans- and greater specific area of the smaller sizes. All the particle
formation from droplet to wet particle. At this stage, since the sizes undergo four distinct slopes of the moisture-loss pro-
rate of moisture removal from the surface becomes depend- files, representing the three drying stages. These are more
ent on the internal diffusion of moisture to the surface, this clearly visible in the moisture profiles of medium size par-
results in a decrease in the drying rate and hence most of the ticles (200–400 ␮m). In the first stage, the moisture-loss slope
heat absorbed causes rapid rise in the particle temperature. along the height is almost linear, corresponding to the surface
For smaller particles, the temperature quickly reaches the gas drying (first stage). The rate of moisture-loss then becomes
temperature and follows the gas temperature profile along the slower. This corresponds to the diffusion controlled drying
tower height afterwards. This is because the smaller parti- stage (second stage). The rate of moisture loss then increases
cles have greater specific surface area, smaller diffusion path again, corresponding to the third stage of drying. The rate
and larger heat and mass transfer coefficients (Fig. 10(a) and of moisture-loss decreases as the particle moisture content
(b) resulting in greater heat and mass exchange rates. From approaches zero since the driving force in this stage is the tem-
the temperature profile of 400 ␮m particle size, it can be seen perature difference between the particle and the gas which
that the temperature remains fairly constant at about 373 K becomes smaller. The particle sizes up to 400 ␮m lose all the
up to a certain height before it starts to rise again. This tem- moisture before exiting from the tower.
perature corresponds to the slurry boiling temperature. Once Fig. 11(a) and (b) is the plots of heat and mass transfer
the particle reaches the slurry boiling point, the drying rate coefficients, respectively, of droplets/particles of various sizes.
becomes dependent on the rate of heat transfer to the particles The initial heat and mass transfer coefficients of the smallest
(third drying stage). The slurry boiling point increases rapidly particle are greatest because they vary inversely with diame-
ter. A sharp decrease in the heat and mass transfer coefficients
occur due to the decrease in the velocity of droplets/particles.
Temperature (K)
300 350 400 450 500 550 600
The minimum velocity of the particles is set to the terminal
0 falling velocity, there onwards the particles fall with a constant
0.1 velocity (see Fig. 14); therefore, the heat and mass transfer
Dimensionless Height (z/Z)

50 µm
0.2 coefficients remain constant.
100 µm
0.3 Fig. 12 is a plot of drying rates of selected particle sizes.
200 µm
0.4 The horizontal axis is plotted on a logarithmic scale due to the
400 µm
0.5 orders of magnitude difference in the drying rate of particles of
800 µm
0.6 different sizes. The drying rate profiles show the three stages
1000 µm
0.7 Gas
of drying and are more clearly visible for larger particle sizes.
0.8 For all the sizes, the drying rate is the highest in the beginning
0.9
of the first stage of drying. The drying rate in the first stage falls
down quickly as the velocity of the injected droplets starts to
1
reduce due to drag, this reduces the mass transfer coefficient
Fig. 9 – Temperature profiles of droplets/particles and hot and hence the drying rate. The drying rate in the second stage
gas. is independent of the surrounding conditions in the tower as it
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 837

(a) Heat Transfer Coefficient (W/m2 K) Drying Rate (g/s)

0 1000 2000 3000 4000 5000 1E-07 1E-06 1E-05 0.0001 0.001
0 0
50 µm 50 µm
0.1 0.1
100 µm 100 µm
Dimensionless Height (z/Z)

0.2 0.2

Dimensionless Height (z/Z)

200 µm
0.3 200 µm
400 µm 0.3
0.4 400 µm
800 µm 0.4
0.5 800 µm
1000 µm 0.5
0.6 1000 µm
0.6
0.7
0.7
0.8
0.8
0.9
0.9
1
1

Mass Transfer Coefficient (m/s) Fig. 12 – Drying rate of droplets/particles.

(b) 0 1 2 3 4 5
0
0.1 50 µm Particle Density (kg/m3)
0.2 100 µm 1100 1200 1300 1400 1500 1600
Dimensionless Height (z/Z)

200 µm 0
0.3
400 µm 0.1 50 µm
0.4

Dimensionless Height (z/Z)

800 µm 0.2 100 µm
0.5 0.3
1000 µm 200 µm
0.6 0.4 400 µm
0.7 0.5 800 µm
0.8 0.6 1000 µm
0.9 0.7
1 0.8
0.9
Fig. 11 – (a) Heat transfer coefficient of droplets/particles, (b) 1
mass transfer coefficient of droplets/particles.
Fig. 13 – Density profiles of droplets/particles.

is internal moisture diffusion controlled, hence it is relatively some stage and therefore the drying rate fluctuates between
uniform. The drying rate increases in the third stage due to the the second and third stages.
internal vapourisation of moisture that facilitates the trans- Fig. 13 is a plot of density of the droplets/particles of
port of moisture to the surface, hence the heat transfer to the selected particle sizes. The density of the droplets in the initial
particle controls the rate of drying. In the third stage of drying, drying stage is nearly constant, since the droplet diameter also
for larger particle sizes, sharp changes in the drying rate can reduces in the first stage due to loss of moisture from the sur-
be seen, this occurs when the particle temperature starts to face. Thereafter, the density profiles of the droplets/particles
rise more rapidly, resulting in a decrease in the heat transfer are very similar to the moisture content profiles in Fig. 10 due
driving force (temperature difference between the particle and to strong dependency of density on the moisture content. The
the gas), the heat transfer to the particle becomes insufficient size of the particle is changed to the initial droplet size in
to maintain the particle at its corresponding boiling point at the third stage of drying to account for particle inflation; this

Dimensionless Velocity (vp /vp,0 )

0 0.2 0.4 0.6 0.8 1
0
0.1 50 µm
Dimensionless Height (z/Z)

0.2 100 µm
0.3 200 µm
0.4 400 µm
0.5 800 µm
0.6 1000 µm
0.7
0.8
0.9
1
Fig. 14 – Velocity profiles of droplets/particles.
838 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

40 0.9
550
0.8
35

Final Moisture Fraction (wl/wl,0)

0.7

Final Temperature (K)

30 500
0.6
Residence Time (s)

25 0.5
450 0.4
20
0.3
15
400 0.2
10 0.1
5 350 0
50 150 250 350 450 550 650 750 850 950
0 Particle Size (µm)
50 150 250 350 450 550 650 750 850 950
Particle Diameter (µm) Fig. 17 – Particle temperature and moisture content.

Fig. 15 – Residence times of particles.

the particles are completely dried, they quickly acquire the
surrounding gas temperature. A sharp decrease in the tem-
results in a sharp decrease in the density of the particles which perature of the particles sizes ranging from 425 ␮m to 650 ␮m
is more clearly visible in profiles of particle sizes 400 ␮m and can be seen. The exit moisture content in this range of parti-
above. The minimum density of the particles exiting at zero cles also shows a sharp rise with increase in particle size. Since
moisture content is about 1120 kg/m3 . The larger particles exit the particles exit at the slurry boiling temperature (third stage
at a greater moisture content hence they have a higher density. of drying), in this stage, the boiling point of slurry is a function
Fig. 14 is a plot of dimensionless velocity profiles (vp /vp,0 ) of moisture content. The slurry boiling point increases expo-
of selected particle trajectories. The initial velocity of all the nentially with decrease in the moisture content as moisture
injected droplets is the same. The smaller particles have becomes more bound. Hence a sharp decrease in tempera-
lower weight; hence the smaller particles lose the momen- ture of the particles is seen as the moisture content increases.
tum quicker and reach the terminal falling velocity at a shorter The exit temperatures of particle sizes greater than 650 ␮m are
distance from the injection location. Once the particles reach fairly constant as the slurry boiling point does not vary sig-
the terminal velocity, the particles continue to fall with this nificantly at higher moisture content. These distributions are
velocity until they reach the bottom of the tower. The termi- used to generate the weighted average values of temperature
nal falling velocity of smaller particles is lower compared to and moisture content.
larger particles. Therefore the smaller particles have greater Case 2: In Case 2, the measured size distribution of the dried
residence times, which are plotted in Fig. 15. powder is used as the initial size distribution of the droplets.
Fig. 16 is a plot of dimensionless velocity of the hot gas Fig. 18 is a plot of temperature profiles of the droplets/particles
along the dimensionless height of the column. The gas veloc- of selected sizes and the hot gas. The temperature profiles of
ity is highest at the bottom due to highest temperature at the the droplets/particles and the hot gas is qualitatively similar
inlet and hence lowest density. As the gas travels upwards, to that obtained from Case 1, however, the range of sizes is
due to heat exchange with the droplets/particles, the temper- different. In the previous case, all the particles exited at the
ature of the hot gas drops and the density increases which boiling point of the slurry. While in this case, the larger par-
decreases the gas velocity. ticles exit at a temperature less than the slurry boiling point.
Fig. 17 is a plot of temperature and moisture content of the This can be seen for 1500 ␮m and 2300 ␮m in the plot. Table 4
particles at the exit of the spray drying tower. Smaller parti- lists the surface drying time and residence time of particles of
cles (up to 450 ␮m) have zero moisture content because these different sizes. The surface drying time of smaller particles is
particles have higher residence times and higher specific area, very short compared to larger particle sizes since the surface
resulting in greater heat and mass transfer. The larger parti- drying time is inversely proportional to the square of the mass
cles exit at a higher moisture content. Smaller particle sizes transfer coefficient, which is greater for smaller particles. For
(up to 450 ␮m) exit at nearly a constant temperature. The mois- larger particles, the residence time is only slightly greater than
ture content of these particle sizes is also zero. Hence once
Temperature (K)
300 350 400 450 500 550 600
Normalized Gas Velocity (vgas /vgas,max) 0
0.7 0.75 0.8 0.85 0.9 0.95 1 100 µm
0.1
0 400 µm
Dimensionless Height (z/Z)

0.2
0.1 800 µm
Dimensionless Height (z/Z)

0.3
0.2 1000 µm
0.4
0.3 1500 µm
0.4 0.5 2300 µm
0.5 0.6 Gas
0.6 0.7
0.7 0.8
0.8 0.9
0.9 1
1
Fig. 18 – Temperature profiles of droplets/particles and hot
Fig. 16 – Dimensionless velocity of the hot gas. gas.
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 839

Final Moisture Fraction (wl /wl,0)

550
0.9
500 0.8

Final Temperature (K)

0.7
450 0.6
0.5
400 0.4
0.3
350 0.2
0.1
300 0
100 500 900 1300 1700 2100
Particle Size (µm)
Fig. 19 – Particle temperature and moisture content.

the surface drying time; hence these particles exit soon after 1 after a dimensionless height of 0.1. Since most of the heat
a crust is formed at the surface. and mass exchange between the droplets and particles with
Fig. 19 is a plot of exit temperature and moisture content the gas takes place in the top region of the tower in Case 1
of the particles as a function of size. The smaller particles due to smaller droplets/particles having greater surface area,
(up to 450 ␮m) exit at almost uniform temperature and the allowing rapid evaporation of moisture. Hence the tempera-
moisture content of these particles is also zero. The particle ture of the gas is greater in the bottom region of the tower due
sizes in the range of 500 ␮m to 650 ␮m show similar behaviour to comparably less heat exchange as most of the heat is taken
as in Case 1, in which the greatest change in moisture con- up in evaporating the moisture from the droplets/particles in
tent and exit temperature is observed with increasing particle the top region of the tower.
sizes from 425 ␮m to 650 ␮m. The particle sizes in the range Table 5 lists the calculated mass weighted average par-
of 700 ␮m to 1200 ␮m exit at almost uniform temperature, ticle temperature and moisture content for the whole size
because the slurry boiling temperature shows very small vari- distribution, the gas outlet temperature and heat loss for the
ation in this range of moisture content. For the particle sizes simulated cases along with the measured values. The mass
greater than 1200 ␮m the temperature starts to fall and the weighted average dried powder moisture content predicted
moisture content is nearly uniform, because these particles by Case 1 is smaller than Case 2 and the measurement. This
exit in the second stage of drying, in which the drying rate is is primarily because of the use of measured droplet size dis-
controlled by moisture diffusion to the surface. The temper- tribution used to represent the droplets/particles. The drying
ature of these particles depends on how much exposure the rates of the droplets/particles vary significantly with the size
particles get with the hot gas. For this range of particle sizes of the droplets/particles. Smaller droplets/particles lose the
(1250 ␮m to 2300 ␮m), since the larger particles have shorter moisture content quicker due to higher specific area. The res-
residence times, hence the exit temperature is also smaller idence times of the smaller droplets/particles are also greater.
compared to the smaller particles. This is the primary reason for over prediction the drying of
Fig. 20 is a plot of heat flux along the column height. The particles in Case 1. Since the dried particle moisture content
negative sign indicates that the heat is lost from the tower
to the surrounding. The heat loss is greatest at the bottom
of the tower because the temperature difference between the
hot gas and the ambient is the largest, as the temperature Table 5 – Simulation results and measured data.
of the hot gas is lowered; the heat loss also becomes rela- Parameter Case 1 Case 2 Experiment
tively smaller. From the comparison of the two cases, it can
Dried powder 1.08 4.62 2.49
be seen that, for a dimensionless height of 0–0.1 from the top,
average
the heat flux in Case 1 is smaller. It becomes greater in Case moisture
content %
(w/w)
Table 4 – Surface drying time and residence time of Powder 525.31 486.12 361.7
selected particle sizes. average
outlet
Particle size Surface Residence
temperature
(␮m) drying time time (s)
(s) (K)
Gas outlet 378.28 402.92 382.3
100 0.012 37.540 temperature
400 0.059 3.094 (K)
800 0.124 1.192 Average 1134 1190 703
1000 0.157 0.906 density
1500 0.241 0.611 (kg/m3 )
2300 0.378 0.455 Heat loss (W) 3426 3265 21,435
840 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841

Heat Flux(W/m 2)
-100 -80 -60 -40 -20
0
0.1
Dimensionless Height (z/Z) 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9 Case 1
Case 2
1
Fig. 20 – Heat flux along the column height.

is smaller, the predicted gas outlet temperature is also smaller to obtain the heat and mass transfer coefficients are applica-
compared to the measurement. ble to the smooth spherical particles, the change in shape of
In Case 2, the dried particle size distribution is used to rep- the particles may also cause a deviation of the heat and mass
resent the initial size distribution of the droplets. This results transfer coefficients from the ones calculated using Ranz and
in under-predicting the drying of particles, since larger par- Marshall (1952) correlations.
ticles exchange less heat with the gas due to lower specific A large difference between the measured powder outlet
area and shorter contact time with the gas in the spray drying temperature and the simulation results can be seen. This
tower. Since the moisture content in Case 2 is greater, hence is mainly because the powder temperature measurement is
the predicted gas outlet temperature is also larger compared taken at the belt which is at a few metres away from the
to the measurement. position where the particles fall from the tower. Before the par-
Other reasons for the differences between the measure- ticles fall on to the belt, the particles come into contact with
ments and the simulation results could be due to the an entrained cold air stream. This is expected to decrease the
simplifying assumptions in the model to predict the par- temperature of the dried powder quickly due to a high surface
ticle residence times including the limitation of minimum area of contact and low heat capacity of the dried powder. The
particle velocity to the terminal falling velocity, the use of predicted heat loss is smaller than the heat loss based on the
smooth particle drag law, no particle deposition on the wall measured value, since the simulation predicts the heat loss up
and no particle-particle interaction. The smaller particles will to the nozzle height. The height of the tower is greater than the
get entrained by the gas, in addition the smaller particles nozzle height hence the heat will be lost from the top region
may get caught up in the recirculating zones, and this will of the tower above the nozzle location. Other reasons could
increase the residence times of the smaller particles. The drag be the imperfect insulation causing greater heat loss from the
law which is used for the calculation of the drag force on the tower and the temperature measurement errors.
particles is valid for smooth spherical particles. However, the
particles will undergo change in shape due to drying as well 7. Conclusions
as agglomeration. The irregular shaped particles may exhibit
larger drag force and hence greater residence time, which is A one-dimensional plug flow model has been developed to
not accounted in the simulation. The wet particles may get simulate droplet/particle drying in a counter-current arrange-
deposited on the wall; the particles may retain at the wall for ment for gas and particle flow in a spray drying tower.
some time and lose moisture before the particles eventually The particles are assumed to be uniformly dispersed over
get entrained back into the gas stream. The particle-particle the cross-section of the tower. The particles size strongly
interaction is also ignored in the simulation. The larger par- influences the temperature and moisture history of the
ticles move faster compared to smaller particles, hence the particles as it affects the residence time of the particles
movement of larger particles will be hindered by the slower and the surface area available for heat and mass transfer.
moving particles and vice versa. These assumptions need to be Smaller droplets/particles have greater heat and mass transfer
relaxed for a better estimation of the particle residence times. coefficients and larger residence times hence these particles
The particles are assumed to move only in the axial direc- exit at lower moisture content. The slurry inlet temperature
tion and the swirl component of the gas phase and the is higher than the wet bulb temperature; therefore, the tem-
particles is ignored. Due to the presence of swirl, most of perature of the droplets begins to fall. However, a constant wet
the particles move close to the wall hence a temperature bulb temperature is not reached since the solids content in the
gradient in the gas phase will be present along the cross- slurry is very high and a crust is formed before a constant wet
section, i.e., a high temperature gas in the central region and bulb temperature can be reached. Thereafter, the temperature
relatively lower temperature gas near the walls. Hence the of the particle begins to rise. A comparison of the simulation
plug-flow assumption may give over-prediction of the spray results with experimental data shows that the case with mea-
drying tower performance. The correlations which are used sured droplet size distribution over-predicts, while the case
 chemical engineering research and design 9 2 ( 2 0 1 4 ) 826–841 841

with dried powder size distribution under-predicts the heat Hecht, J.P., 2012. Personal Communication. P&G Technical
and mass exchange between the gas and the particles. The Centres, UK.
large difference in the results of the two simulated cases indi- Hecht, J.P., King, C.J., 2000a. Spray drying: influence of developing
drop morphology on drying rates and retention of volatile
cate the importance of accurate initial size distribution of the
substances. 1. Single-drop experiments. Ind. Eng. Chem. 39,
droplets as well as inclusion of coalescence/agglomeration to 1756–1765.
allow for changes in droplet/particle diameters along the col- Hecht, J.P., King, C.J., 2000b. Spray drying: influence of developing
umn height. This simple plug-flow model has an advantage of drop morphology on drying rates and retention of volatile
being computationally efficient compared to a more detailed substances. 2: Modeling. Ind. Eng. Chem. 39, 1766–1774.
modelling approach. The model can be used to quickly esti- Kadja, M., Bergeles, G., 2003. Modelling of slurry droplet drying.
mate the performance of a counter-current spray drying tower Appl. Thermal Eng. 23, 829–844.
Kreith, F., 1973. Principles of Heat Transfer, third ed. Intext Press,
with a low computing requirement (few minutes on a desktop
Singapore.
computer). Lin, J.C., Gentry, J.W., 2003. Spray drying drop morphology:
experimental study. Aerosol Sci. Technol. 37,
Acknowledgements 15–32.
Masters, K., 1972. Spray Drying: An introduction to Principles
Partial financial support by Procter and Gamble, Newcastle Operational Practice and Applications. Leonard Hill Books,
London.
Innovation Centre and from IPSE, University of Leeds for the
Mezhericher, M., Levy, A., Borde, I., 2007. Theoretical drying
first author is gratefully acknowledged. The authors would like model of single droplets containing insoluble or dissolved
to thank Mr. Zaeed Alam, Procter and Gamble, for his support solids. Drying Technol. 25, 1035–1042.
and encouragement. Mezhericher, M., Levy, A., Borde, I., 2008. Heat and mass transfer
of single droplet/wet particle drying. Chem. Eng. Sci. 63,
12–23.
References
Montazer-Rahmati, M.M., Ghafele-Bashi, S.H., 2007. Improved
differential modeling and performance simulation of slurry
Audu, T.O.K., Jeffreys, G.V., 1975. The drying of drops of spray dryers as verified by industrial data. Drying Technol. 25
particulate slurries. Trans. Inst. Chem. Eng. 53, 165–172. (9), 1451–1462.
Charlesworth, D.H., Marshall, W.R., 1960. Evaporation from drops Morsi, S.A., Alexander, A.J., 1972. An investigation of particle
containing dissolved solids. AIChE 6, 9–23. trajectories in two-phase flow systems. J. Fluid Mech. 55 (2),
Cheong, H.W., Jeffreys, G.V., Mumford, C.J., 1986. A receeding 193–208.
interface model for the drying of slurry droplets. AIChE 32, Nelson, P.A., Stevens, W.F., 1961. Size distribution of droplets from
1334–1346. centrifugal spray nozzles. AlChE J. 7 (1), 80–86.
Crank, J., 1975. The Mathematics of Diffusion, second ed. Nesic, S., Vodnik, J., 1991. Kinetics of droplet evaporation. Chem.
Clarendon Press, Oxford. Eng. Sci. 46, 527–537.
Dalmaz, N., Ozbelge, H.O., Eraslan, A.N., Uludag, Y., 2007. Heat Parti, M., Palancz, B., 1974. Mathematical model for spray drying.
and mass transfer mechanisms in drying of a suspension Chem. Eng. Sci. 29, 355–362.
droplet: a new computational model. Drying Technol. 25 (2), Perry, R.H., Green, D.W., 1997. Perry’s Chemical Engineers’
391–400. Handbook, seventh ed. McGraw-Hill, New York.
Effting, C., Folgueras, M.V., Guths, Saulo, Alarcon, O.E., 2010. Ranz, W.E., Marshall, W.R., 1952. Evaporation from drops. Chem.
Microstructural characterization of ceramic floor tiles with Eng. Prog. 48, 141–146, 173-180.
the incorporation of wastes from ceramic tile industries. Rosin, P., Rammler, E., 1933. The laws governing the fineness of
Mater. Res. 13 (3), 319–323. powdered coal. J. Inst. Fuel 7, 29–36.
Elperin, T., Krasovitov, B., 1995. Evaporation of liquid droplets Sano, Y., Keey, R.B., 1982. The drying of a spherical particle
containing small solid particles. Int. J. Heat Mass Transfer 38 containing colloidal material into a hollow sphere. Chem. Eng.
(12), 2259–2617. Sci. 37, 881–889.
Farid, M., 2003. A new approach to modeling of single droplet Sunkel, J.M., King, C.J., 1993. Influence of the development of
drying. Chem. Eng. Sci. 58, 2985–2993. particle morphology upon rates of loss of volatile solutes
Furuta, T., Tsujimoto, S., Okazaki, M., Toei, R., 1983. Effect of during drying of drops. Ind. Eng. Chem. Res. 32,
drying on retention of ethanol in maltodextrin solution 2357–2364.
during drying of a single droplet. Drying Technol. 3, 311–327. Topar, J., 1980. Mathematical model of spray drying reckoning
Griffith, J.D., Bayly, A.E., Johns, M.L., 2008. Magnetic resonance with droplet size distribution. In: Drying ‘80: Proceedings of
studies of detergent drop drying. Chem. Eng. Sci. 63, the second international drying symposium, vol. 2,
3449–3456. pp. 405–409.
Handscomb, C.S., Kraft, M., Bayly, A.E., 2009. A new model for the Walton, D.E., Mumford, C.J., 1999. The morphology of spray-dried
drying of droplets containing suspended solids after shell particles: the effect of process variables upon morphology of
formation. Chem. Eng. Sci. 64, 228–246. spray-dried particles. Trans. IChemE 77 (A), 442–460.
Hassan, H.M., Mumford, C.J., 1993. Mechanisms of drying of skin Williams, A., 1976. Fundamentals of oil combustion. Prog. Energy
forming materials. 1. Droplets of materials which gelatinised Combust. Sci. 2, 167–179.
at high temperature. Drying Technol. 11, 1730–1750.