You are on page 1of 11

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/222573470

A comparative study of a spray dryer with


rotary disc atomizer and pressure nozzle using
computational fluid dynamic simulations

Article in Chemical Engineering and Processing June 2006


DOI: 10.1016/j.cep.2005.11.004

CITATIONS READS

56 729

3 authors, including:

A. S. Mujumdar
McGill University
704 PUBLICATIONS 9,321 CITATIONS

SEE PROFILE

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

Numerical investigation of impinging drying with intermittent input View project

All content following this page was uploaded by A. S. Mujumdar on 18 November 2015.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Chemical Engineering and Processing 45 (2006) 461470

A comparative study of a spray dryer with rotary disc atomizer and


pressure nozzle using computational fluid dynamic simulations
Li Xin Huang a, , Kurichi Kumar b , A.S. Mujumdar c
a Department of Equipment R&D, Research Institute of Chemical Industry of Forest Products, Nanjing 210042, PR China
b CFD Division, Institute of High Performance Computing, Singapore 117528, Singapore
c Mechanical Engineering Department, National University of Singapore, Singapore 119260, Singapore

Received 27 July 2004; received in revised form 3 November 2005; accepted 3 November 2005
Available online 28 December 2005

Abstract
Spray drying operations with rotary disc atomizers as well as pressure nozzles as atomizing devices are widely used in diverse industries. The
design of spray dryers is typically optimized for the design conditions. However, users sometimes need to use the same spray dryer chamber with
some modifications due to the requirements of different products and or production rates. In this paper, we present results of a computational fluid
dynamic study carried out to investigate the possibility of multi-functional applications of a specific spray dryer chamber. The predicted airflow
pattern is validated by favorable comparison with published data. The airflow pattern and temperature distributions predicted by the model at
different levels in the drying chamber are presented and discussed. The effects of different atomizer designs are also investigated. Note that the
two types of atomizers yield very different droplet size distributions as well as spray patterns. The volumetric evaporation rate values, heat transfer
intensity and thermal energy consumption per unit evaporation are computed and compared for drying of a 42.5% solids maltodextrin suspension
in a spray chamber 2.2 m in diameter with a cylindrical top section 2.0 m high and a bottom cone 1.7 m high. Wall regions most susceptible to
formation of undesirable deposits are also identified. Computed droplet trajectories for the two spray patterns are shown to display very different
flow, temperature and drying characteristics.
2005 Elsevier B.V. All rights reserved.

Keywords: Atomization; Drying; Flow pattern; Heat and mass transfer; Turbulence model

1. Introduction Full-scale spray dryer simulations using computational fluid


dynamics (CFD) technology is one possible solution to this prob-
Over the past decade, research in the field of spray dryer mod- lem. However, the lack of experimental data in the public domain
eling has primarily dealt with the practical utility of spray dryers adds to the uncertainty when building enhanced physical models
in various industries [13]. These designs are mainly empiri- representative of real spray dryers, which is a major impediment
cal in nature. However, experiments on full-scale spray dryer to innovation in spray drying area.
present major difficulties, not only because of their large sizes Crowe et al. [4,5] proposed the Particle-Source-In-Cell
and massive costs involved, but also because of the complex and model (PSI-Cell) to model the spray drying process. This is
hostile environment in which to measure flow, temperature and an axi-symmetric model and two-way coupling between two
humidity, etc. within the drying chamber. On the other hand, it is phases, i.e., drying medium and spray droplets. Paradakis and
essential to understand the spray drying process well. This will King [6,7] applied this model to a spray drying process. A
lead to a good productivity, low energy consumption and high good comparison between numerical model and experimental
final product quality. results was reported. Oakley and Bahu [8] investigated and
validated the PSI-Cell model in spray drying simulation using
CFD code FLOW3D. Negiz et al. [9] developed a steady state
mathematical model for a co-current spray dryer based on the
work of Parti and Palancz [10].
Corresponding author.
E-mail addresses: l x huang@163.com (L.X. Huang), Langrish and Zbincinski [11] used the CFX code to investi-
kkumar@ihpc.nus.edu.sg (K. Kumar), mpeasm@nus.edu.sg (A.S. Mujumdar). gate the effects of swirl on droplet deposits within a small-scale

0255-2701/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2005.11.004
462 L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470

spray dryer. Kieviet et al. [12,13] investigated a co-current spray equation and the species conservation equation. Two-way cou-
dryer fitted with a pressure nozzle both in CFD simulation and pling between drying medium and droplet is considered as well.
experiments. The CFX code was used. More recently, Huang et The general equation for an axial symmetrical flow in cylin-
al. [14] investigated the effects of different chamber geometries, drical coordinates can be written as:
i.e., cylinder-on-cone, lantern, hour-glass and pure cone, on the
(U) 1 (rV)
drying performance and particle residence time using FLUENT +
code. They showed that it was possible for the designer to select x r r
   
other chamber geometries, and not only the traditional cylinder- 1
= + r + S + Sp (1)
on-cone geometry. Huang et al. [15] also carried out a parametric x r r r r
study for a co-current spray dryer fitted with pressure nozzle. A
ultrasonic nozzle spray dryer was studied numerically by Huang where S is the source term of gas phase, Sp the source term of
et al. [16], as well. Goula and Adamopoulos [17] also used FLU- the droplet phase and the effective viscosity, , is summarized
ENT to simulate a lab-scale spray dryer. in Table 1.
However, previous workers focused attention on spray dryers The turbulence viscosity T is computed by:
fitted with nozzles, in particular, the pressure nozzle atomizers.
k2
Few studies on spray dryers with rotary disc atomizers can T = C g (2)
be found in the literature; most of them are experimental in
nature. Huang et al. [18] showed that the RNG k different The generation of turbulence, Mk , is given as follows:
turbulence model is suitable for simulating the complex,     
  2 
swirling two-phase flow in the spray drying chamber fitted U V 2 U 2 V 2 V
with a rotating disc when compared with three other turbulence Mk = T + +2 + +
r x x r r
models, i.e., standard k, realizable k and Reynolds stress
    
models using FLUENT code [19]. W 2 W W 2
The objective of this work is to present and discuss a set of + + (3)
x r r
numerical results obtained using the CFD code Fluent 6.1 for a
co-current flow spray dryer fitted with either a rotary disk or a
pressure nozzle [19]. A fully three-dimensional configuration, The empirical constants of the turbulence model are taken
i.e., cylinder-on-cone geometry, is considered. The RNG k from the standard constants given by Launder and Spalding [20].
different turbulence model was selected in this study based on The combined EulerLagrangian approach is used to obtain
our previous work [18]. Comparison with limited experimental the particle trajectories by solving the force balance equation for
data is included as well. the particles considering the discrete phase inertia, aerodynamic
drag, gravity gi and other optional user-defined forces Fxi .
2. The mathematical model dupi 18 Re g
= CD (ui upi ) + gi + Fxi (4)
dt p dp 24 g
The hybrid Eulerian and Lagrangian approach is used to
model the heat and mass transfer between two-phase flow (air and the relative Reynolds number
and droplet) in spray drying process. This computation involves
numerical solution of the discretized continuity equation, the dp |up u|
Re = (5)
momentum equations with suitable turbulence model, energy

Table 1
Terms in the general equations
Equation S Sp
dmp
Continuity 1 0 0

 U  1  V
 P
dt
Axial momentum U + T + r + RU
x x r r x x

 U
 1
 V  P g W 2 2V
Radial momentum V + T + r + RV 2
x r r r r r r r
2
Tangential momentum W + T (rW) + RW
T r r
Turbulence energy k + Mk g

kT
Dissipation rate of turbulence kinetic energy + (C1 Mk C2 g )
k
T
Enthapy n + Rn
n
T dmp
Vapor mass faction H + RH
H dt
L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470 463

and drag coefficient Table 2


Boundary conditions used for simulation
a2 a3
CD = a1 + + 2 (6) Air mass flow rate (kg/s) 0.336
Re Re Air temperature ( C) 195
Air absolute humidity (kg/kg) 0.014
The user-defined forces Fxi are set equal zero since the cohe- Spray rate (kg/s) 0.0139
sion effects are not considered in this model. It is also due to the RosinRammler parameter 2.05
difficulty in describing the cohesion phenomena. Feed temperature ( C) 27
The heat transfer between the droplet and the hot gas is com- Air radial velocity (m/s) 5.25
puted according to the heat balance equation as follows: Air axial velocity (m/s) 7.50
Air total velocity (m/s) 9.15
dTp dmp Turbulence k-value (m2 /s2 ) 0.027
mp cp = hAp (T Tp ) + hfg (7) Turbulence -value (m2 /s3 ) 0.37
dt dt Pressure at outlet (Pa) 100.0

The mass transfer between the gas and the droplet are calcu- Chamber wall thickness (m) 0.002
lated according to the following equation: Wall material Steel
Wall-heat transfer coefficient (W/m2 K) 3.5
Air temperature outside wall ( C) 27
dmp
= kc Ap (Cs C ) (8) Interaction B.C. between wall and droplet Escapea
dt a Indicates that the particles are lost from the calculation domain at the point
The heat and mass transfer coefficient can be obtained from of impact with the wall.
Eqs. (9a) and (9b) with suitable coefficient values for , , , ,
and . These values are obtained from several sources [21,22].
The FLUENT code solves the governing equations listed 3. Boundary conditions and atomization models
here.
3.1. Boundary conditions
Nu = 2 + Re Pr (9a)
The simulations were performed for steady state operation.
Sh = 2 + Re Sc (9b) The grid-independence of the results was demonstrated using
different mesh sizes, i.e., 0.03, 0.015 and 0.0075 mm. Finally,
The second period, called droplet boiling, is applied to predict 0.015 mm mesh size is selected in this paper. The chamber is
the convective boiling of a discrete phase droplet when the tem- a cylinder-on-cone vessel, 2.215 m in diameter with a cylindri-
perature of the droplet has reached the boiling point and while cal top section, 2.0 m high and a bottom cone, 1.725 m high.
the mass of the droplet exceeds the non-volatile fraction. The The angle of the bottom cone is 60 . The atomizer is installed
boiling rate equation is applied [FLUENT [19]]: at the top of the drying chamber. Air is blown into it from the

top center as shown in Fig. 1. The feed is a 42.5% solids mal-
d(dp ) 4k Cp, (T Tp )
= (1 + 0.23 Red ) ln 1 + todextrin suspension. The boundary conditions at the inlet, the
dt p cp, dp hfg outlet, chamber wall and the turbulence model are summarized
(10) in Table 2.

Fig. 1. Physical geometry (all dimension in mm) Kieviet [13].


464 L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470

3.2. Atomisation models

Spray from pressure nozzle: the velocity at discharge from


the nozzle is 59 m/s [12,13]. The droplet size distribution is
selected to be such that 10.0 m is the minimum droplet diam-
eter and 138.0 m is the maximum droplet diameter with an
average droplet diameter Dm of 70.5 m; it is modeled using
a RosinRammler distribution with these parameters and the
spread parameter equal to 2.05 [13].
The spray from the rotary disc atomizer: 16 injection points
located along the circumferential wheel edge are defined to spec-
ify the spray with a same mean droplet diameter and droplet size
distribution as that from the pressure nozzle so that the com-
parison between these two atomizers can be carried out. The
ejections at 16 points are defined in the same way as the proce-
dure reported by Huang et al. [18]. The disc diameter is taken
as 105 mm with a rotational speed of 20,000 rpm. If Eq. (11)
[23] is used to calculate the Sauter-mean droplet diameter from
a rotating disk atomizer, a mean value of 74.5 m for droplets is
obtained. This is similar in magnitude to 70.5 m from a pres-
sure nozzle. The tangential and radial velocities are calculated
in Masters [23]. They are 109.9 and 7.67 m/s, respectively. Note
that in practice rotary disc and pressure nozzle are selected for
different applications [2,3].

G0.24
d = 1.4 104 L
0.83
(11)
(NdD ) (ndh )0.12

4. Results and discussion

4.1. Comparison between numerical and measured results


in a pressure nozzle spray dryer

To compare the predicted results with Kieveits mea-


sured data, a three-dimensional CFD model is used to sim-
ulate such a spray dryer with and without a pressure noz-
zle. Although a two dimensional axi-symmetric CFD model
was carried out by Kieviet et al. [12,13] and Huang et al.
[15], they showed that reasonable agreement with measure-
ment results is obtained. In the following 3D simulation results,
it will be easily seen that the flow patterns, temperature and
humidity profiles and contours are asymmetric beyond certain
level.

4.2. Comparison of velocity proles between


prediction and measurement in a pressure nozzle Fig. 2. Comparison of velocity validation between prediction and Kieviets mea-
spray dryer surement [13] at no spray condition.

Fig. 2 compares the velocity profiles at different levels


within the chamber between the predicted [present] and mea- of the 1.0 m level. However, the asymmetric velocity profiles at
sured results [13] under no spray condition. Good agreement is 2.0 m level are found in Fig. 2d. It is due to the internal exit bent
obtained considering the complexity of the process and mea- pipe which reduces the area for air to go through at one side of
surement results. It is noted that there is a non-uniform velocity the drying chamber. It may also be because of the turbulent flow
distribution in the core region of the chamber. The highest veloc- in the drying chamber. This indicates that the symmetric CFD
ity magnitude is about 8.0 m/s at the 0.30 m level. The velocity is models in Kieveits [12,13] and Huang et al. [15] assumptions
reduced as the air streams downwards in the chamber. It is also may lead to some inaccuracies in the predicted results for such
found that the air flow patterns are nearly symmetric upstream spray drying geometry.
L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470 465

Fig. 3. Comparison of temperature profiles between prediction and Kieveits Fig. 4. Comparison of humidity profiles between prediction and Kieviets mea-
measurement [13] at spray condition. surement [13] at spray condition.

4.3. Comparison of temperature proles between prediction


and measurement in a pressure nozzle spray dryer

Under the pressure nozzle spray conditions, the predicted


temperature and the measured data by Kieviet [13] at various
levels are shown in Fig. 3. It showed that the predicted results
agree well with the measured results.
The temperature profiles (Fig. 3) show that the temperature in
the central core region of a diameter of 0.25 m are significantly
different at 0.3 and 1.4 m levels. It is due to the intense heat
and mass transfer during the initial contact between the spray
and drying air because of the high relative velocity between
these two phases coupling with large temperature driving force.
Only a minor radial variation of air temperature is found in the
remaining volume except for the region very near the chamber
wall.

4.4. Comparison of humidity proles between prediction


and measurement in a pressure nozzle spray dryer

From the humidity profiles at the 0.3 and 1.4 m levels shown
in Fig. 4, it is also seen that a large variation of humidity occurs
in the central column. This also indicates that rapid drying takes
place in this region. The remaining volume has almost constant
humidity. This result implies that these regions are not effectively
used for drying and so there are opportunities here to improve
the design and operation.
It is also observed that the humidity profiles at 0.3 m level Fig. 5. Comparison of velocity profiles for a spray dryer fitted with pressure
and 1.4 m level along X-axis are symmetric and are asymmet- nozzle and rotary disk.
466 L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470

ric at 1.4 m level along Y-axis. It implies that the evaporation In Fig. 5, it is noted that the velocity profiles are nearly sym-
occurs non-uniformly at the cross-section of the drying cham- metric at 0.3 and 1.4 m levels for Case A. On the other hand,
ber. It also shows that the symmetric model may not give too asymmetric velocity profiles are found in Case B. It is also found
accurate predicted results. that the absolute velocity is higher in Case B at 0.3 m level than
in Case A. It is due to the disc rotating which induce the air high
4.5. Comparison of a spray dryer tted with a pressure swirl. However, when the air passes downward, such as at 1.4 m
nozzle and a rotating disk level, the high velocity zone only slightly expands in Case A.
Case B at 1.4 m level presents a larger high velocity zone than
4.5.1. Comparison of velocity proles and air ow patterns that in Case A.
between a spray dryer tted with a pressure nozzle (Case A) If the axial velocity profiles in Fig. 6 are considered, it is
and rotating disk atomizer (Case B) observed that there is a reverse flow at 0.3 m level in Case B
The velocity magnitude and axial velocity profiles for a spray which is not present in Case A. It is due to the rotating disk which
dryer (SD) fitted with a pressure nozzle (PN) and a rotating pulls the air below the disk upwards. This phenomenon was
disk (DK) are shown in Figs. 5 and 6, respectively. We named also observed by Masters [23] and was named as air pumping
the a spray dryer fitted with PN as Case A and DK as Case B. effects. These reverse flows disappeared rapidly as the drying air
Generally, it is found that the high variation of velocity is located proceeds downwards, e.g., at 0.6 m level (Fig. 6b). However, at
at the center core of diameter of 0.3 m as what we found in no the same time, there is a reverse flow formed near the chamber
spray condition [Fig. 2]. wall as shown in Fig. 6b and c. At the 1.4 m level, these side

Fig. 6. Comparison of axial velocity profiles for a spray dryer fitted with pressure Fig. 7. Comparison of temperature profiles for a spray dryer fitted with pressure
nozzle and rotary disk. nozzle and rotary disk.
L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470 467

reverse flows are more obvious and strong in Cases A and B. higher volumetric effectiveness of the chamber [14]. We also
This shows that air recirculation appears at this level. The radial find that the temperature variation is quite large in the top region
and tangential velocity profiles (not shown here) show that the of the chamber in Case B.
flows are too complex to identify typical flow patterns.
4.5.3. Comparison of humidity proles for Cases A
4.5.2. Comparison of temperature proles for Cases A and B
and B Fig. 9 displays humidity profiles for Cases A and B at different
The temperature profiles for Cases A and B are shown in levels in the drying chamber. From Fig. 9a, it is noted that the
Fig. 7. It is seen that the temperatures in the central core of humidity in the central core of diameter 0.2 at 0.3 m level in
diameter of 0.2 m are almost the same at 0.3 m level in both Case B is lower than that in Case A. However, this humidity is
cases, but radial variations are larger in Case B than those in Case larger in Case B as the radius increases than that in Case A. This
A. It is because the rotating disk provides a horizontal spray. At is mainly due to the radially directed spray for rotating disk and
0.6 m level (Fig. 7b), it shows almost the same conclusions as downward spray for pressure nozzle. Thus intense evaporation
those for the 0.3 m level. At 1.4 m level (Fig. 7c), Case A gives occurs at different locations because of the designs of different
a lower temperature profile in the central core than Case B. It is atomizers. At 0.6 m level (Fig. 9b), the same conclusion can be
due to the evaporation which takes place in this region since the made.
pressure nozzle provides a downward spray. However, Case B However, Case A presents a more uniform humidity profiles
presents a slightly wider region of high temperature than Case at 1.4 m level than in Case B. It is due to the stronger recirculation
A. It is probably because the rotating disk produces a stronger in Case A, which makes the high humidity air at the lower section
swirling flow and makes the high temperature zone to expand. travel upwards. On the other hand, Case B presents lower humid-
Fig. 8 shows the temperature contours in planes YZ and XZ ity in the central core and high humidity outside this core. Such
for Cases A and B. They indicate that the temperature variation is a condition may lead to less evaporation because the droplets
limited within the narrow central region for Case A and occupies always pass through a high humidity and lower temperature dif-
more regions for Case B. This implies that Case B may yield ference zone.

Fig. 8. Comparison of temperature contours for a spray dryer fitted with pressure nozzle and rotary disk.
468 L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470

Table 3
The overall drying performance for nozzle and disc spray dryers
Cases Case A Case B Case C

Percentage of droplets hitting the wall (%)


Outlet 0.76 1.96 1.77
Cylinder 2.42 9.27 10.85
Conical 13.04 7.00 5.68
Ceiling 0.17 0.45 0.83
Internal pipe in chamber 4.30 2.19 2.10
Deposit rate at the walls (kg/h)
Outlet 3.6 9.2 8.3
Cylinder 11.4 43.6 51.0
Conical 61.3 32.9 26.7
Ceiling 0.8 2.1 3.9
Internal pipe in chamber 20.2 10.3 9.9
Outlet temperature ( C) 116 118.5 113
Real evaporation rate (kg/s) 7.8 103 7.6 103 7.9 103
Total heat consumption (W) 24285 24059 25236
Percentage of evaporation rate 97.6 95.2 98.9
Residual moisture (%wt) 3.1 5.0 1.5
Energy consumption per unit 3114 3165 3194
evaporation rate (kJ/kg)

B. It also indicates that the short cylinder-on-cone geometry is


not optimal for a pressure nozzle SD. The probable reason is
that nozzle spray only gives a narrow spray angle. This also
leads to large deposit on the conical wall in Case A. The simu-
lated deposition rates at different walls are also provided in the
table. From it, the same conclusions to the above discussion are
obtained.
On the other hand, if comparison of the evaporation rates
in both Cases A and B is carried out, it shows that the evap-
oration rate is slightly higher in Case A than that in Case B.
It is the same as what we find from the above analysis, i.e.,
strong recirculation may increase the evaporation rate and on
the other hand, smaller evaporation rates will be obtained if the
Fig. 9. Comparison of humidity profiles for a spray dryer fitted with pressure
nozzle and rotary disk. droplets always pass through a high humidity and lower tem-
perature difference region in such a chamber geometry. Smaller
evaporation rates directly lead to higher heat consumption due
4.5.4. Comparison of overall drying performance for Cases to higher outlet temperatures and larger temperature differences
A and B between the wall and the ambient. Higher deposit percentage in
Table 3 summarizes the overall drying performance for Cases Case A directly leads to reduced flow of dried powders exiting
A and B. We find that the deposits on the cylinder wall for Case B from the outlet.
are significantly higher than those in Case A. It is due to the radial The outlet temperature for SD fitted with a PN is 116 C
spray with a high velocity in Case B. Note should be taken that which is 3 C higher than Kieviets measurement [13]. It is prob-
the escape boundary condition makes the total evaporation rate ably because the higher deposit on bend internal pipe and conical
relatively lower since the particles are lost from the calculation wall leads to fewer evaporation rates. If the exit temperature in
domain at the point of impact with the wall under this BC. It Case B is compared with that in Case A, another 3 C higher
directly leads to the larger-diameter cylinder geometry needed temperature is found in Case B. It is directly due to the slight
in SD with DK [2,3,23]. fewer evaporation rate in Case B. In any case, a 3 C variation
Deposits on the top wall are also predicted in both cases. will be difficult to assess experimentally.
Case B is slight heavy due to the air pumping effects. How-
ever, the pressure nozzle SD presents a large deposit on the 4.6. Comparison of a rotating disk spray dryer using
outlet pipe which is internal to the drying chamber. This occurs different droplet size distributions
because the fast spray from the nozzle travels in the same direc-
tion as the inlet air and thus is prone to hit this pipe. The Although identical droplet size distribution is used in Cases
deposit fraction is almost double in Case A compared to Case A and B so that we can compare these two types of spray dryers;
L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470 469

it directly leads to lower evaporation rate for a rotating disk Appendix A. Nomenclature
SD. The droplet size distribution is obtained from a pressure
nozzle. However, the rotating disk atomizer normally gives a a1 , a2 and a3 constants
smaller mean droplet size and a narrower size distribution than Ap surface area of the droplet (m2 )
a pressure nozzle [23]. The evaporation rate may be increased CD drag coefficient
in Case B if the fine mean droplet and narrow size distribution Cs moisture concentration at the droplet surface (mol/m3 )
which is produced by a rotating disk are used. Case C is designed Cg moisture concentration in the bulk gas (mol/m3 )
for this purpose. Case C is different from Case B only in terms Dm diffusion coefficient of vapor in the bulk (m2 /s)
of the droplet size distribution and mean droplet size, i.e., the cp heat capacity of the droplet (J/kg K)
minimum droplet diameter of 10 m and the maximum droplet dp droplet diameter (m)
diameter of 100 m with an average droplet diameter Dm of d Sauter mean droplet diameter (m)
50 m are used in Case C [23]. dh holes diameter or vane height in disk (m)
The predicted results shown in Table 3 confirm the above dD rotating disk diameter (m)
conclusion, i.e., a higher evaporation rate is obtained in Case dmp
dt rate of evaporation (kg/s)
C due to the small mean droplet size and narrow droplet size
Fxi user-defined-function in particle tracking governing
distribution. The evaporation rate is increased about 4% higher in
equation
Case C than that in Case B. Considering the deposit percentage in
gi gravity component (m2 /s)
Cases B and C, we find that only the deposits on the cylinder wall
Gk production of turbulence kinetic energy due to the mean
are increased significantly. It is probably because the smaller
velocity gradients
particles are easily carried by the drying medium since there is
G production of turbulence kinetic energy due to buoy-
a strong swirling flow in the upper volume of the chamber.
ancy
Because of the higher evaporation rate in Case C, a lower
GL mass feed flow rate (kg/h)
residual moisture in product is obtained as well compared with
h convective heat transfer coefficient (W/m2 K)
those in Cases A and B. The energy consumption values per
hfg latent heat (J/kg)
unit evaporation are almost the same in Cases A, B and C. It
k turbulent kinetic energy (m2 /s2 )
indicates that the narrower droplet size distribution and smaller
kc mass transfer coefficient (m/s)
mean droplet size in spray drying better the overall drying per-
keff effective thermal conductivity of gas (W/m K)
formance.
kg thermal conductivity of the hot medium (W/m K)
mp mass of the droplet (kg)
5. Conclusions
n number of holes or vanes in disk
N disk rotating speed (rpm)
A three-dimensional computational fluid dynamic model for
Ni molar flux of vapor (mol/m2 s)
pressure nozzle and rotating disc spray dryers was developed and
RU , RV and RW user-defined source term in momentum equation
investigated. Good agreement with limited experimental data is
Rn user-defined source term in energy equation
obtained considering complexity of the system studied. It also
Sk user-defined source term
shows that a three-dimensional model is more suitable for such
S user-defined source term
a spray drying system than a two-dimensional axi-symmetric
t time (s)
model [12,13,15].
T temperature (K)
The results obtained from the 3D CFD model are presented in
Tg gas temperature (K)
terms of the velocity magnitudes, velocity components, temper-
ui gas velocity vector (m/s)
ature profiles, humidity profiles and particle trajectories. Results
up droplet velocity vector (m/s)
show that the flow patterns in a SD are very complex. Pressure
nozzle may lead to a high velocity variation core formed in the
center of the chamber. Large recirculation is also found in a Greek letters
pressure nozzle SD. surface tension (N/m)
It is also observed that a greater proportion of the chamber gas density (kg/m3 )
volume is utilized in a rotary disc atomizer SD than in one fit- p droplet density (kg/m3 )
ted with a pressure nozzle SD. There are some regions that are viscosity (kg/m s)
deposit-free in the chamber near cylinder wall in a pressure noz- energy dissipation rate (m2 /s3 )
zle SD. Larger zones of deposits on cylinder wall are found for k turbulent Prandtl number for k
a rotating disk SD and on internal bend and a conical wall for a turbulent Prandtl number for
pressure nozzle SD. These predicted results show that a larger- t turbulent viscosity
diameter chamber should be used for a SD fitted with DK and a
tall small-diameter one for one fitted with PN. Subscripts
From predicted results, it is found that the narrower droplet g air
size distribution and smaller mean droplet diameter present bet- p droplet
ter overall drying performance. C constant
470 L.X. Huang et al. / Chemical Engineering and Processing 45 (2006) 461470

Non-dimensional number  [11] T.A.G. Langrish, I. Zbincinski, The effects of air inlet geometry and
hd
Nu Nusselt number kp spray cone angle on the wall deposition rate in spray dryers, Trans. I.
Chem. E. 72 (A) (1994) 420430.
Pr Prandtl number of gas (cp /kg )  [12] F.G. Kieviet, P.J.A.M. Kerkhof, in: A.S. Mjumdar (Ed.), Using Compu-
dp |
ug 
up |
Re Reynolds number Re = u
tational Fluid Dynamics to Model Product Quantity in Spray Drying:
  Air Flow Temperature and Humidity Patterns; in Drying96, vol. A,

Sc Schmidt number D Krakow, Poland, 1996, pp. 259266.
 m  [13] F.G. Kieviet, Modeling Quality in Spray Drying, Ph.D. thesis, Endin-
k d
Sh Sherwood number Dc mp hoven University of Technology, the Netherlands, 1997.
[14] L.X. Huang, K. Kumar, A.S. Mujumdar, Use of computational fluid
dynamics to evaluate alternative spray chamber configurations, Drying
References Technol. 21 (2003) 385412.
[15] L.X. Huang, K. Kumar, A.S. Mujumdar, A parametric study of the gas
flow patterns and drying performance of co-current spray dryer: results
[1] I. Filkova, A.S. Mujumdar, Industrial spray drying systems, in: A.S.
of a computational fluid dynamics study, Drying Technol. 21 (2003)
Mujumdar (Ed.), Handbook of Industrial Drying, vol. 1, second ed.,
957978.
Marcel Dekker, Inc., New York, 1995, pp. 263308.
[16] L.X. Huang, K. Kumar, A.S. Mujumdar, Simulation of spray evaporation
[2] L.X. Huang, A.S. Mujumdar, Classification and selection of spray dryers,
using pressure and ultrasonic atomizer- a comparative analysis, Trans.
Chem. Ind. Digest (JulyAugust) (2003) 7584.
Tambov State Tech. Univ. 10 (2004) 83100.
[3] L.X. Huang, A.S. Mujumdar, Design of spray dryers, Chem. Ind. Digest
[17] A.M. Goula, K.G. Adamopoulos, Influence of spray drying conditions
(NovemberDecember) (2003) 95102.
on residue accumulation -simulation using CFD, Drying Technol. 22
[4] C.T. Crowe, M.P. Sharam, D.E. Stock, the particle-source-in-cell (PSI-
(2004) 11071128.
Cell) model for gas-droplet flows, J. Fluid Eng. 9 (1977) 325332.
[18] L.X. Huang, K. Kumar, A.S. Mujumdar, Simulation of a spray dryer
[5] C.T. Crowe, in: A.S. Mujumdar (Ed.), Modeling Spray-air Contact in
fitted with a rotary disk atomizer using a 3D computational fluid dynamic
Spray Drying Systems in Advances in Drying, vol. 1, Hemisphere, New
model, Drying Technol. 22 (6), 14891515.
York, 1980, pp. 6399.
[19] Fluent Manual. Discrete Phase Models, 2004 (Chapter 19;
[6] S.E. Papadakis, J. King, Air temperature and humidity profiles in spray
www.fluent.com).
drying. 1. Features predicted by the particle source in cell model, Ind.
[20] B.E. Launder, D.B. Spalding, Lectures in Mathematical Models of Tur-
Eng. Chem. Res. 27 (1988) 21112116.
bulence, Academic Press, London, England, 1972.
[7] S.E. Papadakis, J. King, Air temperature and humidity profiles in spray
[21] Van der Sanden, C.T. Coumans, P.J.A.M. Kerkhof, in: A.S. Mujum-
drying. 2. Experimental measurements, Ind. Eng. Chem. Res. 27 (1988)
dar (Ed.), Desorption Isotherms and Diffusion Coefficients of Catalytic
21162123.
Materials, Drying98Proceedings of 11th International Symposium on
[8] D.E. Oakley, R.E. Bahu, Computational modeling of spray dryers, Com-
Drying (IDS98), 1998, pp. 762769.
put. Chem. Eng. 17 (1993) 493498.
[22] R.H. Perry, Perrys Chemical Engineers Handbook, seventh ed., McGraw
[9] A. Negiz, E.S. Lagergren, A. Cinar, Mathematical models of cocurrent
Hill, New York, 1997, pp. 70180.
spray drying, Ind. Eng. Chem. Res. 34 (1995) 32893302.
[23] K. Masters, Spray Drying Handbook Longman Scientific & Technical,
[10] M. Parti, B. Palancz, Mathematical model for spray drying, Chem. Eng.
John Wiley & Sons Inc., New York, 1991, pp. 200505.
Sci. 29 (1974) 355362.

View publication stats

You might also like