See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272790516

CFD Simulation of a Co-current Spray Dryer

Article · February 2010

CITATIONS READS

6 900

1 author:

Saad Saleh
Tikrit University
11 PUBLICATIONS   41 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Production of monodisperse prills View project

All content following this page was uploaded by Saad Saleh on 25 February 2015.

The user has requested enhancement of the downloaded file.

 World Academy of Science, Engineering and Technology 62 2010

CFD Simulations of a Co-current Spray Dryer

Saad Nahi Saleh
 dimensional and axisymmetric in order to reduce the demand
Abstract—This paper presents the prediction of air flow, on computational resources. There is clear experimental
humidity and temperature patterns in a co-current pilot plant spray evidence to demonstrate three dimensional behavior in this
dryer fitted with a pressure nozzle using a three dimensional model. type of equipment suggesting that numerical simulations of
The modelling was done with a Computational Fluid Dynamic spray dryers need to include the three-dimensional nature of
package (Fluent 6.3), in which the gas phase is modelled as the flows [7]. Previous two-dimensional, axisymmetric
continuum using the Euler approach and the droplet/ particle phase is
simulations can only be regarded as indicative, at best, since
modelled by the Discrete Phase model (Lagrange approach).Good
agreement was obtained with published experimental data where the
they do not reproduce the basic physics that are involved [8].
CFD simulation correctly predicts a fast downward central flowing As explained above, it is apparent that in order to avoid
core and slow recirculation zones near the walls. In this work, the wall build up and insufficient residence time of particle which
effects of the air flow pattern on droplets trajectories, residence time influences the product quality, we must able to model of the
distribution of droplets and deposition of the droplets on the wall also complexity of the spray/air mixing and including the proper
were investigated where atomizing of maltodextrin solution was drying model , therefore the objective of this paper is to
used. perform two phase simulations of a co-current spray dryer in
order to examine what current computational fluid dynamics
Keywords—Spray, CFD, multiphase, drying, droplet, particle. techniques are capable of achieving when simulating such a
system and to understand what happened inside the spray
I. INTRODUCTION dryer?

S PRAY dryer is an essential unit operation for the

manufacture of many products with specific powder
properties, e.g. chemical, ceramic, food; pharmaceuticals etc. II. MODELLING APPROACH
In spite of the wide uses of the spray dryers, they are still In order to simulate the spray drying process in a rigorous
designed mainly on the basis of experience and pilot way, it is necessary to gain insight into the flow pattern, local
experiment [1]. One of the big problems facing spray dryer temperature and local moisture content of the air and the
designer and operators is the complexity of the spray/air temperature-time history of drying particle.
mixing process in spray chamber [2] where the air flow The first important aspect in modelling of spray dryer is
patterns existing inside the spray dryer is considered as one of prediction of the flow pattern of air which depends on the
the primary factors that influence the residence time of droplet geometry of the dryer and the location and design of the air
/ particle, in turn the equality of the product produced by the inlet and air outlet channel. The trajectories followed by the
dryer such as moisture content, size distribution, and bulk drying particles depend not only on the air flow pattern but
density. The particle residence time and surrounding air also on the position and method of atomization.
The second important aspect is the need to include the
temperature are particularly important in the spray drying of
proper modelling of the drying behaviour of the droplet to
thermal sensitive products, such as milk, where product
ensure that the prediction of the droplet behaviour in the dryer
degradation can occur if the particles remain in an air stream
is correct. The above considerations will be discussed in more
for too long, or experience an air stream is too hot [3]. detail as follows.
A very important phenomenon of spray dryer operability is
the particle wall deposition which is affected by the A. Two Phase Flow Simulation
temperature and humidity patterns inside the dryer when
moist particles contact the spray dryer wall. Such depositions The flow in a spray dryer is turbulent and two-phase (gas
can lead to a build up large amounts of product on the wall. and droplets or gas and particles). There are two commonly
These depositions may be dangerous, as they can fall and used approaches for modelling two-phase flow[9]. Firstly,
cause damage to the chamber walls, or they can char, one can treat the disperse phase as an extra fluid with its own
resulting in a potential explosion hazard [4]. flow field (Euler/Euler approach). In the case of spray drying,
Application of computational fluid dynamic (CFD) with rather concentration of particles, one usually use the
techniques in analyses of spray dryers have been carried out second approach, the Euler/ Lagrange approach. In this
successfully and reported by [5], [6] and others. Most of these approach the gas field is calculated first (Euler). This is done
earlier works assume the flows in the dryers are two- by calculating solutions of the Navier-Stockes and continuity
equation on a grid of control volumes. Subsequently the
Saad Nahi Saleh is with the Chem. Eng. Dept., Tikrit University, Iraq
particles are tracked individually (Lagrange). Along the
(e-mail: saad_nahi68@yahoo.com). particle trajectories the exchange of mass, energy and

772
 World Academy of Science, Engineering and Technology 62 2010

momentum with the continuous phase is calculated. These which would be in equilibrium with the surrounding gas and
transfer terms are added to the source terms of the Navier- can be predicted from correlation of sorption isotherms of the
Stokes equations of the gas flow calculation. After the dried material such as the one proposed by [16].
particle tracking, the air flow calculation pattern is 1m
recalculated, taking the transfer terms into account. This cycle 1 ª ln 1  \  º
x eq « » (5)
of airflow calculation followed by particle tracking is 100 ¬  c1 T  c 2  ¼
repeated until convergence is reached. This scheme is called
the Particle-In-Cell model [10]. Where x eq is the equilibrium moisture content on a dry basis
The droplet field is established by integrating the
(in kg water/kg dry material), T is the temperature of the gas
differential equations for droplet motion to determine droplet
(in K), and \ is the relative humidity of the gas (a fraction
velocities and, with further integration, droplet trajectories. At
each time step along the trajectory, droplet size and from 0 to 1), and c1 , c 2 and m are constants with the values
temperature history are calculated using the equations for 0.000405, -187.962 and 1.169 respectively.
droplet mass and heat transfer rates. These equations can be The characteristic drying curve for a given material is
found in [11]. Since space is limited, they are not repeated unique and independent of external drying conditions.
here. Furthermore, it is hypothesized that the drying-rate curves for
The effect of turbulence on the droplet motion is modelled the same material at different operating conditions will be
by the turbulent stochastic model. Turbulent stochastic geometrically similar, i.e., the normalized drying rate curve
tracking of droplets admits the effect of random velocity characterizes the drying of a given material[17], therefore it
fluctuations of turbulence on droplet dispersion to be can be assumed that the critical moisture content is the same
accounted for in prediction of droplet trajectories [12]. as the initial moisture content.
The concentration of vapour at the droplet surface is
B. Droplet Drying evaluated by Raoult's Law , where the partial pressure of the
vapour at the surface is equal to mole fraction of the water
In this modelling approach, the drying kinetics are
included and the concept of a characteristic drying curve has ( X s ) multiplied by the saturated vapour pressure, Psat at the
been used here [13]. This essentially empirical approach has droplet temperature, Td :
been widely used for the modelling of single particle drying, X s Psat
where it was based on the assumption of two distinct periods C w, s (6)
RTd
of drying, namely, the constant rate period which is then
followed by the falling rate period. The approach relies upon The concentration of vapour in the bulk gas is calculated by:
first identifying an unhindered drying rate which may P
C w, f Y w , f (7)
correspond to the rate in the first/constant drying period, and RTf
may mathematically expressed as Where Yw,f is the local bulk mole fraction of water vapour,
P is the local absolute pressure, and Tf is the local bulk
dm
Nˆ k c Ad M w C w, s  C w,f  (1) temperature in the gas. The mass transfer coefficient in (1) is
dt
calculated from the Sherwood number correlation [18]:
The relative drying rate, [ , is then defined
kc d d
N Sh AB 2.0  0.6 Re d 1 2 Sc 1 3 (8)
[ (2) D i,m
Nˆ
Where N is the drying rate and N̂ is the unhindered drying
rate. [ is a function of the characteristic moisture content, III. CASE STUDY
defined as
For CFD simulation, the spray dryer used in this article is a
[ f )  if x d xcr co-current pilot plant spray dryer by Niro Atomizer as shown
[ 1 if x ! xcr (3) in Fig. 1. The geometry and air inlet size are the same as those
x  xeq used by [6]. The nozzle atomizer is located at the top of the
) drying chamber; hot drying air enters the chamber through an
xcr  xeq
annulus with the nozzle as its centre. The outlet of the spray
The drying curve can be fitted by a simple expression of the dryer is a pipe mounted through the wall of the cone section of
form [14]: the chamber, bent downwards in the centre of the chamber.
n This type of spray dryers is a more complex geometry than the
N § x  xeq ·
[ ) ¨n ¸ (4) simple box configuration, which requires an unstructured
Nˆ ¨ xcr  xeq ¸
© ¹ mesh for accurate representation using 84000 tetrahedral mesh
According to [15],the value of n for maltodextrin is 3.22. elements (Fig. 2). To check whether the solution was
Where x is the volume-averaged moisture content, x cr is the dependent on the mesh which had chosen, the mesh was
refined to 160000 elements. For each element of original
critical moisture content and x eq is the solid moisture content

773
 World Academy of Science, Engineering and Technology 62 2010

mesh, the value of axial, radial and tangential velocities was pure feed water and 50 kg/hr of 42.5 wt% maltodextrin
compared with the corresponding values in refined mesh. solution. The nozzle was a hollow-cone-type centrifugal
pressure nozzle (Spraying Systems Co.: SX-type), spray angle
$
and velocity for pure water are 73 and 49 m s  respectively
$
while for maltodextrin solution are 76 and 59 m s  . A spray
is represented by a 10 droplet sizes, ranging from 10 to 138
Pm .Rossin Rammler distribution parameters (mean diameter
and power ) are ( 68.6 Pm and 2.45) for pure water( 70.5
Pm and 2.09) for maltodextrin solution .

IV. BOUNDARY CONDITIONS

A k  H model was chosen to model the turbulence. The
k  H model is the most commonly used in engineering
practice because it convergence considerably better than the
algebraic stress model (ASM) and Reynolds stress model
(RSM) and require less computational effort [6]. To determine
the fate of the droplet trajectory when hit the wall of the
Fig. 1The geometry of the pilot plant spray dryer (the dimensions in drying chamber, Fluent [11] submit multi options that could
mm) selected for the present work. The possible fates for a droplet
trajectory are as "escaped", "trapped", "evaporated',
The differences were smaller than 4%, therefore we can say "reflected" and "coalesced". In this work, 100 droplet
that solution is mesh independence. As a first step, the trajectories were calculated and the "escaped" boundary
modelling of the air flow without spray and swirl is condition is used, where the droplets are lost from calculation
performed. A second step will be the modeling the at the point of impact with the wall. Of each droplet trajectory,
temperature and humidity when pure water droplets are the time of flight and the location of the end-point were
tracked through the air under operational conditions with 5 calculated. In order to compare the CFD predicted results with
swirl degree. experimental results of [6], the same conditions were used as
tabulated in Table 1.

V. RESULTS AND DISCUSSION

Analysis of the CFD simulation of air velocity profile
without spray in the spray dryer, as shown the contour of air
velocity (Fig. 3), showed that the flow field consists of a fast
flowing downward core with a slow recirculation around that
core near the upper section of the conical part of the chamber.
The core broadens as going down to the outlet. The predicted
and measured velocities at different levels are depicted in Fig.
4. One note that the central core is of the radius of about
0.25m and the reminder of the chamber is at very low velocity
(0.2 m/s). The highest velocity magnitude in the core is about
6.75 m/s at the 0.6 m level. The sharp descent of velocity
magnitude at the axis of the chamber is reduced as the air
Fig. 2 Surface mesh for the pilot plant spray dryer goes into the cone section of the chamber. This trend is
agreed very well with the measured results. Only the
predictions of velocity magnitudes at 1.0 m level are
somewhat higher than the measured values. We expect the
The last stage is modeling the spray drying process of reason is that the air flow reveals periodicity in the velocity
maltodextrin solution in co current spray dryer. For modelling magnitude at several locations in the chamber and the 1.0 m
the air flow without spray, the velocity components (axial, level is one of its. This nature of air flow was noticed by [6]
radial and tangential) of inlet air were: 6.03,-4.22 and 0.0 m/s when measuring the velocity signal in these locations.
respectively. Two feed are used for simulating, 42 kg/hr of Therefore we can say that the air flow in this pilot plant spray

774
 World Academy of Science, Engineering and Technology 62 2010

dryer is transient in nature and we need a transient CFD It is practical to compare the predicted values with the
simulation to consider this behaviour. measured values at certain points like outlet. As shown in Fig.
5 and 6 , the predicted values of temperature and humidity for
matodextrin solution spray at outlet are 115 oC and 0.033
kg w /kg a and this agreed well with measured temperature and
TABLE I humidity at the outlet (the measured values at outlet are 113
BOUNDARY CONDITIONS USED FOR CASE STUDY o
C and 0.031 kg w /kg a ). The predicted values of temperature
Air Air temp. Air Feed and humidity for water spray at outlet are 88 oC and 0.044
flowrate humidity temp. kg w /kg a , and this agreed well with measured temperature and
(m3/s) (o
C) (kg w /kg a ) (o C ) humidity at the outlet (the measured values at outlet are 86
0.42 195 0.009 27 o
C and 0.043 kg w /kg a ).
Air axial Air radial Air Swril
velocity velocity tangential Degree
(m/s) velocity
(m/s) (m/s)
7.42 -5.19 0.649 5
Pressure Turb. - k - Turb. - H - Chamber
at outlet value -value wall
thickness
(Pa) (m2/s2) (m2/s3) (m)
-150 0.027 0.37 0.002
Wall Wall heat Air temp. Wall
material transfer outside boundary
coefficient wall condition
(W/m2.K) (o C )
Steel 3.5 25 escaped

It is important to mentioned that the air flow pattern for the Fig. 4 Predicted and measured velocities at different levels measured
feed water and maltodextrin look very similar to the air flow from the ceiling (0.3, 0.6, 1.0 m) in the spray drying chamber
with no spray (Fig. 3), and this emphasize that effects of
droplet/ particles on the air flow pattern in the spray dryer is
very weak, therefore we can adopted the hypothesis which In Fig. 7, a 100 particle trajectories which represent the
ignoring the effects of droplets on the gas flow pattern if the spray of 10 particle sizes are predicted. It is clearly seen that
feed rate is less than 10% of the gas mass flow rate [19], [20]. the particles have different fates and different flight times
In Fig. 5 and 6, the predicted counters of temperature and (particle residence time) . As can be seen, a bigger particles
humidity of air with maltodextrin spray are depicted. From (greater than 81 Pm ) appear to be able to penetrate the fast
these figures, it can seen that a large volume of the chamber flowing core into the slow recirculation zone whereas the
has almost constant temperature and humidity(120 oC, 0.032 smaller particles are trapped in this core where the massive
kg w /kg a ). It appears that most of the evaporation takes place evaporation occur for this smaller droplets/ particles in the
in the fast flowing core. core region due to high air temperature as shown in contour
of temperature pattern (Fig. 5 ), therefore the bigger particles
have longer residence time than the smaller particle(the mean
residence time of 138 Pm particle is 3.6 second while is 1.85
second for 10 Pm particle), and this is opposite to what
estimated by previous models such as Kerkhof's model [21]
which was based upon gravity effect and neglect the air flow
pattern behaviour inside the spray dryer.

Fig. 3 Contour of predicted air velocity (m/s) in the spray dryer

chamber (without spray)

775
 World Academy of Science, Engineering and Technology 62 2010

Fig. 5 Contour of predicted air temperatures ( oC) distribution in the

spray dryer (with maltodexrin spray) Fig. 7 Predicted droplet trajectories (100 droplet tracks represent a
10 droplet size distribution) in the spray dryer chamber. The colour
It is generally assumed that the residence time distributions of each trajectory indicate the size of droplet (in m) as shown in the
left column.
of the particles/droplets are equal to those of the air (Masters
(14)). With air flow rate 0.421 m3/s, a mean residence time of
air is (24.2 s) in this dryer [6] . Analysis of CFD simulation
shows that mean particles residence time in the dryer (2.4 s) is
shorter than mean air residence time.

Fig. 6 Contour of predicted air humidities (kg w /kg a ) distribution in

the spray dryer (with maltodexrin spray)
Fig. 8 Vector plot of the predicted air velocity (m/s) in the lower
conical of the spray dryer chamber.
It is found that a large fraction of the droplets (45%) hit
the conical part of the chamber wall, 9% of the droplets hit
the cylindrical part of the chamber wall, 25% hit the outside
of the air outlet, 1% of the particles hit the roof and 20% of
particles were dragged along to the cyclone. It is expected
that this large deposition at the conical part is because of the
VI. CONCLUSION
air flow pattern effects in this region where the air approaches
the wall at right angles, the air velocity is zero, and this The CFD simulations correctly predict the internal
location can be considered as stagnation point (Fig. 8) which behaviour of the spray dryer. It discerned that the air flow at
reveal higher deposition. This explanation is consistent with specified conditions consist of fast flowing core and slow
the measured results [22]. recirculation zone around it, and the air flow reveals a
periodicity in some locations in the dryer which can
simulated by a transient CFD model. The drying of droplet
take place in the core region where the smaller droplets
evaporated due to high air temperature. The air flow pattern
have an important effect upon the trajectories of the particles

776
 World Academy of Science, Engineering and Technology 62 2010

that specify the particle history and its fate inside the spray REFERENCES
dryer, and in turn , the air flow pattern play an important role [1] Masters, K., Spray Drying Handbook, 4th ed., John Wiley and Sons,
in existence of the deposition and in the quality of the New York, 1985.
product ,in addition to the performance of the spray dryer. [2] Oakley, D.E., Produce uniform particles by spray drying, Chem. Eng.
The air flow pattern with water or maltodextrin solution Prog., October, pp. 48-54, 1997.
[3] Harvie, D.J.E.; Langrish, T.A.G ;Fletcher D.F., A computational fluid
spray is very similar to that without spray, therefore the effect dynamics study of a tall-form spray dryer,Trans. IChemE, 80(3), pp.
of particles on the air flow can be neglected in the spray dryer. 163-175, 2002.
[4] Harvie, D.J.E.; Langrish, T.A.G.; Fletcher, D.F., Numerical simulations
of gas flow patterns within a tall-form spray dryer, Trans. IChemE,79
(3) A, pp. 235–248, 2001.
NOMENCLATURE [5] Oakley, D.E.; Bahu, R.E.,Computational modeling of spray drying,
Comput. Chem. Eng. 17, pp. 493-498,1993.
Ad droplet surface area m2 [6] Kieviet, F.G.,Modelling Quality in Spray Drying,Ph.D. thesis,
Endinhoven University of Technology, the Netherland, 1997.
C w, s concentration of vapour kmol / m3 [7] Southwell, D.B.; Langrish, T.A.G.,Observations of flow patterns in a
spray dryer, Drying Technology, 18 (3),pp. 661–685, 2000.
DW , m diffusion cofficient m2 / s [8] Langrish, T.A.G.; Fletcher, D.F., Prospects for the Modelling and
Design of Spray Dryers in the 21st Century. Drying Technology, 21 (2),
dd droplet diameter Pm pp. 197–215, 2003.
[9] Sommerfeld, M.,Review on numerical modeling of dispersed two phase
kc mass transfer coefficiet m/ s flows, Proc. 5th Int.Symp. on Refined Flow Modelling and Turbulence
Mw moleculer weight of vapour kg/kmol Measurements, Paris, 1993.
[10] Crowe, C.T.; Sharma, M.P. ; Stock, D.E. The particle-source-in-cell
m droplet mass kg (PSI-Cell) model for gas-droplet flows, Journal of Fluid Engineering,
N drying rate kg/s 99, pp. 325–332, 1977.
[11] Fluent Manual, Chap. 23: Discrete phase models, www.fluent.com,
P pressure pa 2006.
[12] Haung, L.; Kumar, K.; Mujumdar, A. S.,A Pararmeteric Study of the
R universal gas constant ( J K ˜ mol ) Gas Flow Patterns and Drying Performance Of Co- current Spray
T temperature K Dryer, Drying Technology, 21, 6, pp. 957-978, 2003.
[13] Langrish,T.A.G.; Kockel,T.K., The implementation of a characteristic
t time s drying curve for milk powder using a computational fluid dynamics
Re Reynolds number - simulation, Chem. Eng. J., 83 (4), pp. 69-74,2001.
[14] Keey, R. , Drying of Loose and Particulate Matter, 1st edn, Hemisphere
Sc Schmidt number - Publishing Corporation , 1992.
[15] Woo, M.w. ;Daud, W.R.; Mujumdar, A.S.; Talib, M.Z.; Hua,W.Z. and
Sh Sherwood number -
Tarsirin,S. M., Comparative study of droplet drying models for CFD
x moisture content kg w / kg s modeling, Chem.Eng. Res.and Des.,86,pp. 1038-1048,2008.
[16] Truong, V.; Bhandari, B.R.;Howes, T., Optimization of co-current spray
X mole fraction in liquid - drying process of sugar-rich foods. Part I—moisture and glass transition
Y mole fraction in gas - temperature profile during drying, Journal of Food Engineering 71 (1),
pp. 55–65,2005.
Greek letters [17] Handscomb, C., Kraft, M., Bayly, A., Spray Drying Modelling,
University of Cambridge, 2005.
[ relative drying rate -
[18] Ranz, W.; Marshall,W., Evaporation from drops. Chem Eng.
) characteristic moisture content - Prog.,48(141),pp. 173,1952.
[19] Reay, D., Modelling continuous convection dryers for particulate solids-
\ relative humidity - progress and problems, Drying ’85, 5,pp. 67-74, 1985.
subscript [20] Elghobashi, S.,On predicting particle-laden turbulent flows, Applied
Scientific Research, 52,pp. 309-329, 1994.
cr critical [21] Kerkhof, P.J.A.M.; Schoeber,W.J.A.H., Theortical modeling of the
eq equilibrium drying behaviour of droplet in spray dryers, Advances in
preconcentration and dehydration of foods, ed. A. Spicer, Applied
d droplet science publisher, London, pp. 349-397, 1974.
[22] Kota, K.; Langrish, T., Prediction of Deposition Patterns in a Pilot-Scale
s droplet surface Spray Dryer Using CFD Simulations, Chem. Prod. and Proc. Modeling,
sat saturation 2(3), 26, 2007.

w water
f bulk air

777

View publication stats