This article was downloaded by: [North Dakota State University]

On: 16 November 2014, At: 16:09

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK

Drying Technology: An
International Journal
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/ldrt20

USE OF COMPUTATIONAL
FLUID DYNAMICS TECHNIQUES
TO ASSESS DESIGN
ALTERNATIVES FOR THE
PLENUM CHAMBER OF A
SMALL SPRAY DRYER
a a
D. B. Southwell , T. A. G. Langrish & D. F.
a
Fletcher
a
Department of Chemical Engineering , University
of Sydney , NSW, 2006, Australia
Published online: 11 Dec 2006.

To cite this article: D. B. Southwell , T. A. G. Langrish & D. F. Fletcher (2001) USE

OF COMPUTATIONAL FLUID DYNAMICS TECHNIQUES TO ASSESS DESIGN ALTERNATIVES
FOR THE PLENUM CHAMBER OF A SMALL SPRAY DRYER, Drying Technology: An
International Journal, 19:2, 257-268, DOI: 10.1081/DRT-100102902

To link to this article: http://dx.doi.org/10.1081/DRT-100102902

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of
the authors, and are not the views of or endorsed by Taylor & Francis.
 The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in
relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any form
to anyone is expressly forbidden. Terms & Conditions of access and use
can be found at http://www.tandfonline.com/page/terms-and-conditions
Downloaded by [North Dakota State University] at 16:09 16 November 2014
 DRYING TECHNOLOGY, 19(2), 257–268 (2001)

USE OF COMPUTATIONAL FLUID

DYNAMICS TECHNIQUES TO
Downloaded by [North Dakota State University] at 16:09 16 November 2014

ASSESS DESIGN ALTERNATIVES FOR

THE PLENUM CHAMBER OF A
SMALL SPRAY DRYER

D. B. Southwell, T. A. G. Langrish, and D. F. Fletcher

Department of Chemical Engineering,

University of Sydney, NSW 2006, Australia

ABSTRACT

The inlet region of a pilot-scale, co-current spray dryer was

simulated using the proprietary Computational Fluid
Dynamics (CFD) codes, CFX4 and CFX5. Several design
alternatives were considered for correcting uneven inlet air
distribution, which is known to influence spray dryer perform-
ance and airflow patterns. The simulations were used to assess
each alternative prior to construction, assuming isothermal
and incompressible flow conditions. Experimental measure-
ments were compared with the simulation results for the orig-
inal and one modified design.
Drying air is supplied to this dryer via an overhead pipe
feeding an annular plenum chamber, of diameter 400 mm,
surrounding the atomiser. A distributor plate with two con-
centric rings of 50 holes, each of 5 mm diameter, forms the
base of the plenum chamber. A three-dimensional grid was
required to model each of the 100 holes separately and to
consider the asymmetric flow behaviour. The resulting grid

257

Copyright & 2001 by Marcel Dekker, Inc. www.dekker.com

 258 SOUTHWELL, LANGRISH, AND FLETCHER

consisted of about 532,000 cells. The CFD simulations proved

useful in predicting the trends in flow distributions in each of
the designs.

Key Words: CFD; Inlet conditions; Spray drying.

INTRODUCTION
Downloaded by [North Dakota State University] at 16:09 16 November 2014

Inlet conditions are known to significantly influence particle trajec-

tories in a spray dryer. Oakley et al. (1988) reported that flow patterns in
a co-current dryer were particularly sensitive to strong inlet swirl. They also
simulated the inlet in order to estimate velocities, turbulence quantities and
pressures at the inlet to the drying chamber. Asymmetric flow patterns were
also noted by Stafford et al. (1996) in a pilot-scale dryer with a nozzle
atomiser and again these may also be due to a slightly uneven flow distri-
bution from the air inlet. This paper reports the use of CFD as an investi-
gative tool in assessing several design alternatives to overcome the poor flow
distribution in a plenum chamber for a pilot-scale unit.

EXPERIMENTAL APPARATUS

The spray dryer at The Department of Chemical Engineering,

The University of Sydney is a modified Niro unit. The key com-
ponents are shown schematically in Figure 1. It is a co-current dryer
with an internal diameter of 0.80 m, a cylindrical section height of
0.59 m and a conical section height of 0.63 m. Air is fed to the plenum
chamber by an overhead pipe of 60 mm diameter. Not all components
shown are described as the experimental circuit consisted of the inlet fan,
the heater and flow measurement section, ducting and the dryer. A
Yokogawa Model YF100 Vortex Flowmeter (F) is installed downstream
of the heater.
The nature of the modifications to the original design gave the plenum
chamber an approximately triangular cross-section, evident in Figure 1. The
original profile of the lid can also be seen in Figure 1, the plenum chamber
being simply welded over the top and an inlet pipe pierced through. The air
distributor plate, shown in Figure 2 following modification, has two con-
centric rings of 50 holes, each of 5.0 mm diameter, and bolts to the base of
the plenum chamber.
 COMPUTATIONAL FLUID DYNAMICS TECHNIQUES 259

Outlet
Lid Seal Atomizer Second
Blower Tw
Instrument Plenum
ports chamber Td

Cyclone
Tw
Downloaded by [North Dakota State University] at 16:09 16 November 2014

Spray Dryer
Td
T F

Heater Inlet

Bottle T Main
Blower
T

Figure 1. Schematic of dryer rig - not to scale.

Figure 2. Distributor plate following modification.

PROBLEM DESCRIPTION

A ‘‘rule of thumb’’ is that uniform flow distribution (5%) results

if the ratios of the kinetic energy of the inlet stream to pressure drop
 260 SOUTHWELL, LANGRISH, AND FLETCHER

across the outlet hole and of friction loss in the duct to pressure drop across
the outlet hole are  1:10. Contrary to observations, hand calculations
suggested the existing design should perform adequately. Correction was
desirable, although it was unclear whether the uneven inflows would ulti-
mately have a major effect upon overall flow patterns. CFD techniques
appeared to offer a way of evaluating design alternatives before any con-
struction and experimental testing. The alternatives reported in this paper
are the original design and a design with a 3608 dished annular baffle
Downloaded by [North Dakota State University] at 16:09 16 November 2014

located under the inlet pipe to redirect the inlet flow around the plenum
chamber. A secondary flat annular baffle is located at the level of the top of
the dish baffle.

EXPERIMENTAL METHOD

All tests were conducted at ambient temperature, without heating

and using only the inlet blower. Removal of the dryer lid, necessary for
measurement access, eliminates the pressure drop associated with the
remainder of the dryer circuit. Consequently, the inlet fan alone delivered
approximately 125 kg.hr-1 to the test circuit rather than the normal
90 kg.hr-1 expected with the dryer lid in situ. Thus, an inlet air flow-rate
of 125 kg.hr-1 at 298 K was used for all simulations.
The pressure drop through the downstream dryer chamber and
associated piping would ensure an approximately uniform pressure near
the distributor plate. Although a slight positive pressure prevails inside
the dryer with the lid in situ (because this pressure field is approximately
uniform, similar to the atmospheric pressure prevailing when experimental
measurements were taken with the lid removed) it is unlikely that the
distribution trends have been significantly affected. Therefore, measure-
ments and simulations at 125 kg.hr-1 can be assumed to be representative
of the air distribution at 90 kg.hr-1.
Pressure was measured at each hole using a Halstrup digital
manometer Model EMA150 through a 0.3 m silicon tube of 5 mm
diameter. The maximum air speed at each hole was measured using
a hand-held Solomat Multifunction meter Model MFM500C fitted
with a Solomat Model 129-MSX mean-velocity single-wire hot-wire
anemometer probe. The probe was moved over each hole, 5 mm
above the plate, until a stable maximum was found. Measurement of
individual velocity components at each hole was not practical with this
device.
 COMPUTATIONAL FLUID DYNAMICS TECHNIQUES 261

CFD SIMULATIONS AND GRID DESIGN

Considerable difficulty was encountered in obtaining adequate grid

resolution for the 100 holes in the distributor plate while keeping the
required computational resources within realistic limits. The simpler idea
of representing all the holes with an annular patch would not have allowed
prediction of any poor distribution.
A ‘short’ grid was initially used to minimise the required computing
Downloaded by [North Dakota State University] at 16:09 16 November 2014

resources. This grid used a short inlet pipe and represented the flow domain
only as far as the inside face of the distributor plate. Each hole was treated
as a separate exit into free space. Inlet conditions to the plenum chamber
were obtained by separately simulating the bent section of pipe feeding the
plenum chamber assuming an air flow-rate of 125 kg.hr-1. Values for u, v
and w velocities, pressure and turbulence parameters, k and e, were set via
an intermediate file and the User Fortran features of CFX4. The axial
velocity is denoted by v, while the u and w velocity components are cross-
stream ones.
Simulation results for ‘short’ grid simulations either exhibited symp-
toms of poor convergence of results that did not seem physically sensible.
Specifically, the flow through the 100 holes was inconsistent and exhibited
unphysical variations as the flow progressed around the plenum chamber. A
simulation using double precision arithmetic, to ensure that round-off effects
were not important, did not eliminate these problems. These simulations
are not explored further but illustrate the importance of choosing an
appropriate solution domain.
As the ‘short’ grid results were unsatisfactory, each simulation was
repeated using a larger grid. The ‘long’ grid was created (Figure 3) to
assess the significance of a connected downstream body of air. The import-
ant features of this grid are the extended inlet pipe, three-dimensional repre-
sentation of the distributor plate and 100 holes and the translation of the
downstream boundaries of the flow domain to a point well away from the
distributor plate.
A satisfactory solution could only be arrived at by extruding the dis-
tributor plate 0.4 m downstream, filling the centre of the annulus, expanding
the diameter of the downstream flow domain beyond the diameter of
the distributor plate and using pressure boundaries on all surfaces of the
downstream flow domain. The resulting grid consisted of approximately
532,000 cells.
All simulation results reported here are based upon the assumptions of
isothermal and incompressible flow at 298K. Since no swirl was introduced
via the plenum chamber, the k-e turbulence model was used. Acceptable
 262 SOUTHWELL, LANGRISH, AND FLETCHER
Downloaded by [North Dakota State University] at 16:09 16 November 2014

Figure 3. Grid boundaries of ‘long’ grid.

convergence was achieved in all cases. False time stepping proved helpful
with the ‘short’ grid, but was not necessary for the ‘long’ grid simulations.

PRESENTATION OF RESULTS

A single speed value for each hole does not adequately reflect the
velocity profile that exists across each hole. Consequently, calculation of a
mass flow-rate from the predicted and measured speed values will not, and
does not, result in the measured mass flow-rate of 125 kg.hr-1. However,
scaling each set of speed values to give a nominal flow rate of 125 kg.hr-1 is
acceptable because a consistent technique was used to acquire all values
prior to scaling. The effect of scaling is that the trends become of more
 COMPUTATIONAL FLUID DYNAMICS TECHNIQUES 263

interest than the absolute speed values and comparisons can be made
more easily.
Pressure values have not been scaled but are reported as variations
from the value found at hole 1 of each row of holes. While this point was
chosen arbitrarily, it has been used consistently as a reference point for all
simulation and experimental results. Reporting pressure variations instead
of absolute values ensures that all values are expressed in a comparable
manner.
Downloaded by [North Dakota State University] at 16:09 16 November 2014

Line plots of average speed, pressure variation and u, v and w veloc-

ities around the plenum chamber provided a convenient way of assessing
results. Holes are numbered 1 to 50 for both the inner and outer rows. The
inlet pipe was located above the holes around the 25th inner and outer pair.
An evenly distributed flow would display near constant, and non-zero,
average speed, v velocity and pressure variation values together with near-
zero u and w velocity values for all 100 holes. A design providing such an
outcome would be regarded as successful.

SIMULATION AND EXPERIMENTAL RESULTS

For the original plenum chamber, scaled average speed values and
pressure variations for the simulation and experiment are presented in
Figure 4. Trends, rather than local fluctuations, are of most interest. The
predicted peak in average speed for both rows of holes corresponds to the
point directly below the inlet pipe and was observed experimentally.
However, the predicted peak was not as pronounced as that observed. A
local pressure rise is also observed directly below the inlet pipe, which is
attributable to the holes in this region being subject to a greater air velocity
than holes further away from the inlet.
It is evident from the results presented that the pressure and velocity
profile distributions around the distributor plate are not as smooth as would
be expected physically. This is a consequence of the difficulty of obtaining
well-converged solutions using CFX4 for this complex, multi-block geome-
try. It is well known that for large meshes and a multi-block geometry that
the SIMPLE algorithm is hard to converge because of the relatively weak
coupling between the velocity and pressure fields in this segregated solution
approach. The way around this problem is to use a coupled solver, such as
CFX5, in which the pressure and velocity fields are solved simultaneously
using a multigrid method. Details of this approach can be found in
Hutchinson, Galpin and Raithby (1988). CFX5 has the added benefit that
it uses an unstructured mesh, so that the flow can be solved simultaneously
throughout the flow domain, rather than on a block by block basis.
 264 SOUTHWELL, LANGRISH, AND FLETCHER

Original Design - Scaled CFX4 Speed Profiles

24.00
22.00
Speed (m .s -1)

20.00
18.00
16.00
Downloaded by [North Dakota State University] at 16:09 16 November 2014

14.00
12.00
10.00
0 10 20 30 40 50
Hole #

Experiment - Inner Experiment - Outer Model - Inner Model - Outer

Original Design - CFX4 Pressure Variation Profiles

60.00
50.00
Pressure Variation (Pa)

40.00
30.00
20.00
10.00
0.00
-10.00 0 10 20 30 40 50
-20.00
-30.00
Hole #
Inner - Experiment Outer - Experiment Inner - Model Outer - Model

Figure 4. Original design experimental and predicted results (CFX4) – Scaled speed
and pressure variations (not scaled).

In the CFX5 calculations the same mesh as that used in the

CFX4 calculations was employed. The mesh import routine auto-
matically converts the hexahedral mesh into a tetrahedral mesh. At
present it is not possible to model thin surfaces in CFX5, therefore we
examined only the case with no baffles present. These simulations
converged very well and produced a much more highly converged
 COMPUTATIONAL FLUID DYNAMICS TECHNIQUES 265

Original Design - Scaled CFX5 Speed Profiles

24.00
22.00
Speed (m .s -1)

20.00
18.00
16.00
Downloaded by [North Dakota State University] at 16:09 16 November 2014

14.00
12.00
10.00
0 10 20 30 40 50
Hole #

Experiment - Inner Experiment - Outer Model - Inner Model - Outer

Original Design - CFX5 Pressure Variation Profiles

60.00
50.00
Pressure Variation (Pa)

40.00
30.00
20.00
10.00
0.00
-10.00 0 10 20 30 40 50
-20.00
-30.00
Hole #
Inner - Experiment Outer - Experiment Inner - Model Outer - Model

Figure 5. Original design experimental and predicted results (CFX5) – Scaled speed
and pressure variations (not scaled).

solution. The CFX5 pressure and velocity profiles, shown in Figure 5,

are smoother and more physically realistic than those predicted by CFX4.
Therefore, the use of a coupled solver results in a demonstrable improve-
ment in solution quality when compared with the earlier CFX4 results for
this problem.
 266 SOUTHWELL, LANGRISH, AND FLETCHER

RESULTS FOR MODIFIED PLENUM CHAMBER

Evaluation of each of the simulation results suggested that a 3608

conical baffle might be suitable. This design was adopted with the addition
of three low annular walls located either side of each ring of holes to better
align the flow through each hole with the axis by reducing the other velocity
components. It was necessary to represent each wall as ‘thin surface’ patch
along cell walls in the model. Therefore, because the annular walls could not
Downloaded by [North Dakota State University] at 16:09 16 November 2014

be represented in a totally realistic manner, the results must be seen as only

indicative. The scaled average speed values and pressure variations (not
scaled) are presented in Figure 6.
The baffles have eliminated the region of relatively high pressure
directly below the inlet pipe. These results show flatter profiles than those
presented for the original design. Agreement between the experimental and
predicted values for scaled speed profiles is quite good and the two sets of
values agree better than the same results for the original design. There are
local fluctuations in both profiles but, because their magnitudes are small in
relation to the average speed, these can be overlooked in favour of their
overall trends. The predicted inner speed profile exhibits greater fluctuation
than the predicted outer speed profile, a feature not observed for simulations
without annular walls. As annular walls were only used in the final design,
difficulty in representing these walls in a totally realistic manner, particularly
the innermost wall, is likely to be responsible.
The predicted pressure variation profiles show peaks near hole 10
similar to those predicted for the original design. This is a characteristic
feature of the predicted pressure variation profiles for every design studied,
the peak in the outer row of holes always being slightly greater than that for
the inner row of holes. These peaks, which were not observed experiment-
ally, are considered to be due to unavoidable grid distortion. Although the
pressure variation profiles are not in good agreement with the experimental
values for approximately 40 percent of the holes, they are generally flatter
than the corresponding profiles for the original design. The simulation
results for the modified design represent an improvement in the flow distri-
bution over the original design.

CONCLUSIONS

Although the CFD simulations did not predict the experimental

observations exactly, CFD techniques were successfully used to evaluate
several alternative designs for a practical problem. A design that CFX4
predicted would perform well formed the basis of modifications and per-
 COMPUTATIONAL FLUID DYNAMICS TECHNIQUES 267

Modified Design - Scaled CFX4 Speed Profiles

24.00
22.00
Speed (m .s -1)

20.00
18.00
16.00
Downloaded by [North Dakota State University] at 16:09 16 November 2014

14.00
12.00
10.00
0 10 20 30 40 50
Hole #

Experiment - Inner Experiment - Outer Model - Inner Model - Outer

Modified Design - CFX4 Pressure Variation Profiles

60.00
50.00
Pressure Variation (Pa)

40.00
30.00
20.00
10.00
0.00
-10.00 0 10 20 30 40 50
-20.00
-30.00
Hole #
Inner - Experiment Outer - Experiment Inner - Model Outer - Model

Figure 6. Modified design experimental and predicted results – Scaled speed and
pressure variations (not scaled).

formed well when tested. Use of such techniques may offer potential savings
of time and/or money.
Extension of the flow domain downstream of the distributor plate
resulted in more realistic solutions. This implies that, for a problem such
as this one, extension of the flow domain downstream is necessary.
 268 SOUTHWELL, LANGRISH, AND FLETCHER

Preliminary work with CFX5, an unstructured solver using a coupled solver,

has indicated that improved results, particularly a smoother pressure profile,
can be obtained. It is concluded that CFD techniques were of practical
assistance in assessing design alternatives for inlet plenum chambers in
spray dryers.

REFERENCES
Downloaded by [North Dakota State University] at 16:09 16 November 2014

1. CFX4 (1998), CFX4.2 Solver manual, AEA Technology, Harwell,

Didcot, UK.
2. CFX5 (1998), CFX5.3 User Guide, AEA Technology, Harwell, Didcot,
UK.
3. Hutchinson, B.R., Galpin, P.F., and Raithby, G.D., ‘‘Application of
additive correction multigrid to the coupled fluid flow equations.’’
Numerical Heat Transfer, 13, 133-147, 1988.
4. Oakley, D.E., Bahu, R.E. and Reay, D., (1988), ‘‘The aerodynamics of
co-current spray dryers’’, Sixth International Drying Symposium IDS ‘88,
Versailles, France.
5. Stafford, R.A., Fauroux, O. and Glass, D.H., (1996), ‘‘Flow visualisation
and instantaneous velocity measurements of spray dryer gas and spray
flows using particle image velocimetry’’, Drying 96, Proceedings of the
10th International Drying Symposium (IDS ‘96), Krakow, Poland, A:
555-562.