You are on page 1of 53

This article was downloaded by: [Moskow State Univ Bibliote]

On: 10 October 2013, At: 04:32


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

MODELLING OF PARTICULATE DRYING IN THEORY AND


PRACTICE
a a
Ian C. Kemp & David E. Oakley
a
SPS , Hyprotech , AEA Technology , Gemini Building, Harwell, Didcot, Oxfordshire OX11
0QR, UK
Published online: 06 Feb 2007.

To cite this article: Ian C. Kemp & David E. Oakley (2002) MODELLING OF PARTICULATE DRYING IN THEORY AND PRACTICE,
Drying Technology: An International Journal, 20:9, 1699-1750, DOI: 10.1081/DRT-120015410

To link to this article: http://dx.doi.org/10.1081/DRT-120015410

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
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

DRYING TECHNOLOGY
Vol. 20, No. 9, pp. 1699–1750, 2002

MODELLING OF PARTICULATE DRYING


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

IN THEORY AND PRACTICE

Ian C. Kemp* and David E. Oakley

SPS, Hyprotech, AEA Technology, Gemini Building,


Harwell, Didcot, Oxfordshire OX11 0QR, UK

ABSTRACT

The range of theoretical models used by SPS for the design


and analysis of dryers, especially dispersion dryers, is
reviewed. Different levels of complexity are appropriate at
different stages of the design process. Models should use an
appropriate level of rigour for the available data and the
required purpose of the results. Scale-up from experimental
design curves often yields better results in practice than a
highly complex model requiring many parameters, and the
type of drying kinetics measurements required also depends
on the type of model to be used. Models at four different
levels are presented in general terms and for a number of
major types of dryers. Significant developments in the SPS
methodology for fluidised bed, pneumatic conveying and
cascading rotary dryers are noted. It is also important to
consider the dryer in the context of the overall process
flowsheet.

*Corresponding author. E-mail: ian.kemp@aspentech.com

1699

DOI: 10.1081/DRT-120015410 0737-3937 (Print); 1532-2300 (Online)


Copyright & 2002 by Marcel Dekker, Inc. www.dekker.com
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1700 KEMP AND OAKLEY

Key Words: Heat and mass balance; Scoping design;


Performance; Scale-up; Integral model; Incremental model;
Fluidised beds; Rotary dryers; Kinetics

INTRODUCTION

It is a great pleasure for us to submit this review paper in tribute to


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Professor Roger Keey’s work in drying for over 30 years, and in apprecia-
tion of the major contributions to drying theory and its practical application
by him, his students and colleagues.
Roger Keey has always recognised the importance of rigorous theo-
retical analysis, and his textbooks and monographs bear witness to his
mastery of this. However, he has also sought how best to apply the
theory in practice, aiming for ‘‘fitness for purpose’’ and an appropriate
level of theoretical rigour for the data available. This is especially important
in a field like solids drying, where many of the controlling physical
parameters are not available from databanks and cannot be measured
easily or accurately. Thus, he has also helped to develop and encourage
the widespread use of simple yet effective concepts, such as the characteristic
drying curve in drying kinetics, and accurate experimental measurement
of drying properties. These methods have proved extremely useful in analys-
ing drying data from real industrial situations, focusing on the basic trends
of what is happening, without becoming bogged down in detail. Yet the
rigorous theory is still available when needed, for example if a refinement to
the simple techniques is suggested, or there is a query about their applic-
ability in certain circumstances. This combination of intellectual rigour and
pragmatism has been especially valuable for the work of SPS (Separation
Processes Service), which since 1974 has performed technology transfer and
development of improved selection, design, modelling and optimisation
procedures for dryers and other separation equipment. We are pleased to
have worked closely with Professor Keey throughout the last 25 years and to
acknowledge his very valuable contribution to our work. In this review
paper, the results, both directly and by implication, of his work can be
seen underpinning the current framework of methods for design and
modelling of industrial dryers.
The first comprehensive review of the SPS approach to dryer model-
ling and its application on various specific dryer types was given by Reay
(1989). Twelve years later, there have been advances in several areas, while in
others the approach has stood the test of time with only minor improvements.
Various papers have been published on specific developments, but this is an
appropriate time to update the review and to critically evaluate the success
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1701

or otherwise of the various proposed models, both on theoretical grounds


and as applied in industrial practice.

PREAMBLE: INHERENT PROBLEMS


IN APPLYING THEORY

Drying could be described as the graveyard of academic theory.


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

The subject is intellectually complex, involving simultaneous heat, mass and


momentum transfer processes, giving a highly non-linear set of governing
equations. Numerous parameters affect drying processes, many of them
highly dependent on solid structure. Many of these parameters are difficult
to measure; many vary for the same material made by different processes, or
even between different batches from the same process! This contrasts with
liquid- and vapour-phase processes, where the system is controlled by
equilibrium thermodynamics and all necessary physical properties can be
obtained from databanks. It is a major reason for the oft-quoted statistic
that 90% of fluid processes have reached their design capacity within a year
of initial commissioning; for any process involving solids, the figure is 67%.
So one-third of dryers never reach their expected performance! Moreover,
this statistic was recently re-evaluated and there had been no major
improvement in the last 15 years.
Drying appears deceptively simple to outside observers, because it is
an everyday process that everyone can imagine—drying clothes on the
washing line or in the tumble dryer, drying your hair after a shower,
drying stacks of hay or cereal in the sun. It is easy to imagine the basic
drying process, to understand a little bit about it and to do it in a basic form.
But it is very difficult to do drying really well, to understand it in depth or to
optimise it rigorously. This is why this apparently simple process has a long
track record of theoretical models whose predictions have been wildly dif-
ferent from actual results, and equipment that has failed to meet its design
specification or has been heavily overdesigned to ensure that it can achieve
its scheduled duty.
No drying theory can be considered to be proven and reliable until it
has been experimentally validated, preferably under a wide range of condi-
tions and on both large and small scale. Regrettably, many of the highly
detailed and erudite theories published over the years have included little or
no comparison with practical results on real dryers.
There has historically been a great gulf between drying theory and
industrial drying practice. Manufacturers have tended to rely on empirical
scale-up rules based on pilot-plant testing, rather than published theoretical
models. To some extent this was because the theory was rarely presented in
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1702 KEMP AND OAKLEY

the form of a clear step-by-step design procedure. However, the very limited
acceptance of academic theories for drying by manufacturers and industrial
users has also been due to bad experiences when attempts were made to use
published theory for practical design. This is starkly illustrated by examin-
ing the results from one design procedure published in a textbook (Williams-
Gardner, 1971). He gives the following equation for estimating the drying
time  S and hence duct length L (¼ UG S, where UG is gas velocity) of a
pneumatic conveying (flash) dryer:
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

P dP X  d 2 X
S ¼ ¼ P P ð1Þ
6hPG Tlm 12G Tlm

For the SPS pilot-plant pneumatic conveying dryer, which is 7.1 m long,
the experimentally measured moisture removal X is 0.083 kg kg1 for
silica gel particles of diameter dP ¼ 1.5 mm and density P ¼ 1200 kg m3,
with conveying air at 150 C inlet temperature and 15 m s1 inlet velocity.
Substituting this value of X into Eq. (1) gives a calculated residence time
of approximately 40 s and a dryer length of 600 m, nearly 100 times too high!
Why the discrepancy? There are no actual errors in the derivation,
which is based on a heat balance around the particle. However, it makes
a number of incorrect assumptions;
(a) The heat transfer coefficient used is that for stagnant gas round a
sphere, corresponding to a Nusselt number of 2, whereas it can be
five times greater in the initial acceleration zone where there is a
high relative velocity and flow field.
(b) It uses the log mean temperature difference instead of local
temperature differences. Driving forces are extremely high near
the feedpoint, heat transfer is rapid and the particles heat up
quickly. The gas and particles are close to thermal equilibrium
at the outlet and the driving force is low, but effective drying can
still continue, based on the heat already contained in the gas and
solids (evaporative cooling).
(c) The real particle velocity UP is significantly less than the gas
velocity UG, especially in the initial acceleration zone, and this
effect is enhanced for large particles.
(d) No allowance is made for falling-rate drying kinetics.
Williams-Gardner correctly gave a warning in his book that ‘‘the
procedure is highly theoretical and would be unreliable for design purposes
unless supported by pilot-scale tests’’, but it is doubtful that he envisaged
dryers nearly half a mile high. It can be seen why dryer manufacturers have
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1703

been reluctant to use published theory for designing industrial dryers,


especially as they have to give performance guarantees!
Nevertheless, theory can give good results, if an appropriate level of
rigour is applied. A procedure similar to Williams-Gardner’s, but based on
localised conditions and moving along the dryer (an incremental model),
has been advocated by several workers and applied in practice by SPS—see
below. This procedure gave predictions of duct length and outlet moisture
content within 30% of the actual value, based on fundamental physical
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

property data alone. The error is still significant because of the large
number of parameters required by the model, some of which are difficult
to measure, and for which the errors can be cumulative. When experimental
results from pilot-plant tests were used to ‘‘tune’’ the model and give a better
fit for uncertain parameters, predictions within 10% were obtained on
scaling to different dryer dimensions and different operating conditions.
It is regrettable that a number of over-simplified and dubious theories
continue to be referenced in the literature. For example, the correlations for
particle movement in rotary dryers by Friedman and Marshall (1949) and
Saeman and Mitchell (1954) have been used in recent papers, although
better (and not over-complex) models have been available for many years.
Of course, the age of a theory is no bar to its applicability, as was demon-
strated when Kemp, Bahu and Oakley (1990, 1991) performed experimental
studies on heat transfer to single particles in flash dryers and compared the
results with a wide range of proposed correlations. The best fit by far was
given by the classic Ranz and Marshall equation of 1952, beating a host of
more recently published correlations:

hPG dP
Nu ¼ ¼ 2 þ 0:6Re0:5 Pr0:33 ð2Þ
G

However, the success of this equation is not coincidental; it has a firm


basis in theory (dating back even further, to Froessling in 1938). The con-
stant term corresponds to heat transfer to a sphere in stagnant gas and the
second term allows for the gas flow, boundary layer and wake. Note that
this equation form is not one which would be predicted by any standard
computer curve fitting package, which will tend to go for a straightforward
polynomial, exponential or logarithmic fit. The experimental validation of
the other suggested correlations had been done over too narrow a range of
conditions. If the correct form of the correlation has not been deduced, it will
obviously be highly unreliable when extrapolated. The message is clear—it is
essential to understand the physical significance of the processes that you are
modelling, and also to validate your theory by experiments over a wide
enough range of conditions.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1704 KEMP AND OAKLEY

TYPES OF DRYER MODEL

The approach taken by SPS has been based on dividing dryer models
into two parts; an equipment model and a material model (Reay, 1989)
(Fig. 1). The equipment model includes factors which depend on the type
of dryer used; particle transport through the dryer, external heat transfer
from the hot gas or hot surfaces to the solids, and vapour-phase mass
transfer. These are all external to the particle, and can be modelled theor-
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

etically for the various different types of dryer. The material model covers
factors dependent on the nature of the solid being dried; drying kinetics,
equilibrium moisture content relationships, product quality (e.g. trace com-
ponents) and materials handling. Some of these properties can be found
from databanks, but the majority are highly dependent on the nature and
structure of the solid and on the upstream particle formation process, and
must be measured experimentally (materials characterisation).
Dryer models can be categorised in several different ways:
. The mode of calculation; design, performance or scale-up
. The complexity of the model
. The physical processes considered; heat and mass balance, heat
and mass transfer, drying kinetics and equilibria.
. The level of information generated; basic heat and mass balance,
approximate sizing, overall dryer dimensions, localised conditions
in the dryer.
Many different types of dryer model are available, ranging from the
very simple to the highly complex. It is helpful to consider the different types
of drying calculation under three criteria; the aim or purpose, the physical
processes considered, and the level of complexity.

Figure 1. Representation of the overall drying model as equipment and material


models.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1705

1. The Purpose of the Calculation

The ultimate aim of any drying model is to have a useful practical


application to real industrial-scale dryers. Three particular types of calcula-
tion can be considered:
a. Design of a new dryer to perform a given duty.
b. For an existing dryer, calculation of performance under a different
set of operating conditions.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

c. Scale-up from laboratory-scale or pilot-plant experiments to a


full-scale dryer.
For solids drying, a particular difficulty arises because the internal
moisture transport within a solid is highly dependent on the internal struc-
ture. The latter varies with the upstream process, the solids formation step
and often between individual batches, so that many drying parameters
within solids (e.g. diffusion coefficients) cannot be predicted from theory
alone, or obtained from physical property databanks; practical measure-
ments are required. Hence experimental work is almost always necessary
in order to design a dryer accurately, and scale-up calculations are more
reliable than design based only on thermodynamic data. The experiment
is used to verify the theoretical model and find the difficult-to-measure
parameters; the full-scale dryer can then be modelled accurately.

2. The Physical Processes Modelled in the Calculation

A heat and mass balance around a dryer is straightforward in principle


(although in practice, measurements of even basic parameters such as solids
flowrate and inlet and outlet content may be highly inaccurate).
The initial stages of drying (the induction, or warm-up period, and the
unhindered, or constant-rate period) are controlled by vapour-phase heat and
mass transfer between the hot gas (or heated surface) and the solids. This can
usually be evaluated with reasonable accuracy from correlations. The heat
and mass transfer and the particle movement can be highly dependent on the
air flow patterns, especially in the so-called dispersion dryers where each
individual particle is surrounded by a continuous phase of gas. This is most
marked of all in dryers where the particles are entrained in a moving gas flow,
such as flash and spray (atomisation) dryers. The equipment model is the
main effect on performance during these periods.
The later stages of drying (the hindered, or falling-rate period) are
dominated by internal moisture transport, which is a complex combination
of competing processes; diffusion and convection in both liquid and vapour
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1706 KEMP AND OAKLEY

phases, capillary action in pores, adsorption and chemisorption. The material


model (the nature of the solids being dried) is the main factor. Theory alone
cannot predict how a given material will behave, and experimental measure-
ment of drying kinetics and equilibria will be necessary. Thus, the falling-rate
period is by far the most difficult period to model accurately. Unfortunately,
since drying rates are lowest in this period, it tends to have the dominant effect
on the required drying time.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

3. The Level of Complexity of the Calculation

Modelling can be carried out at different levels, depending on the


amount of data available and the level of precision required.
Simple heat and mass balances give some useful information, but say
nothing about the required equipment size or the performance which a given
dryer is capable of.
Scoping or approximate calculations give rough sizes and throughputs
(mass flowrates) for dryers, using simple data and making some assumptions.
Scaling calculations give overall dimensions and performance
figures for dryers by scaling up drying curves from small-scale or pilot-plant
experiments.
Detailed methods are needed to gain an accurate picture of what is
going on inside a dryer and to track the local conditions of the solids (and gas)
as drying progresses. Naturally, these methods require a lot more input data
and much more complex modelling techniques.
Combining the three sets of criteria above, we can summarise the
characteristics of the available types of dryer models (Table 1). Scoping
methods rely on heat and mass balances and external (vapour phase) heat
and mass transfer. Scaling methods and detailed design models make some
allowance for falling-rate drying kinetics as well, but therefore almost
invariably require some experimental measurements.
At the first three levels, gas and solids flows through the dryer are
considered only in terms of mean inlet and outlet conditions (total mass
or volume flowrate, mean gas velocity, average inlet and outlet humidity).
At the detailed design level, however, localised conditions within the dryer
must be understood.

The Limiting Factor for Dryer Performance

Gardiner (1996) and Kemp and Gardiner (2001), in describing


dryer troubleshooting methods, categorised dryer performance problems
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Table 1. Comparison of Features of Available Models for Dryers

Type of Physical Basis of Dryer Details


Calculation Calculation Calculated Main Uncertainties Typical Use

Heat and mass Heat and mass None Mass flowrate, Performance
balance balance only moisture content,
gas humidity
Scoping design; (a) Heat and mass Overall dimensions: (b) Heat transfer Design
continuous balance only, (a) Cross-section/ coefficients, falling (scale-up?)
convective dryers (b) heat transfer diameter, (b) length rate drying kinetics
Scoping design Heat and mass Overall dimensions; Ht transfer coefficients, Design
for continuous balance, heat diameter, length, temperature difference, (scale-up?)
contact dryers transfer residence time falling rate kinetics
MODELLING OF PARTICULATE DRYING

Scoping design Mass and volume Overall dimensions; Drying time and all Design
for batch dryers of solids diameter, length factors influencing it (scale-up?)
Scaling (integral H&MB, heat Overall dimensions; Local variations Scale-up,
model) transfer, drying diameter, length, inside dryer performance
curve residence time
Detailed design; H&MB, heat Length (use scoping 3-Dimensional flows, Design,
incremental transfer, full method for D); parameter scale-up
model (1-D) drying kinetics local conditions measurement
Detailed design; H&MB, heat Local conditions Measuring required Design, scale-up,
CFD transfer, kinetics, throughout dryer parameters, performance
3-D flow patterns computing time
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016

1707
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1708 KEMP AND OAKLEY

in three ways:
– heat and mass balance deficiencies (not enough heat input to do the
evaporation),
– drying kinetics (drying too slowly, or solids residence time in dryer
too short),
– equilibrium moisture limitations (reaching a limiting value, or
regaining moisture in storage).
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

In the broader sense, even at the design stage or for a successfully


operating dryer, it is useful to know whether its overall performance and
maximum throughput are limited by the heat and mass balance, kinetics
or equilibrium moisture content. Use of this approach for debottlenecking
convective dryers is again described by Kemp and Gardiner (2001). For the
heat and mass balance, the main factors are:
. solids throughput
. inlet and outlet moisture content
. temperatures and heat supply rate
. leaks and heat losses.
Drying kinetics, which are affected by temperature, particle size, and
structure, are limited by external heat and mass transfer to and from the
particle surface in the early stages, but internal moisture transport is the
main parameter at lower moisture. Equilibrium moisture content increases
with higher relative humidity, or with lower temperature.
For example, if it can be shown that drying kinetics is the limiting
factor for a particular dryer, because a minimum solids residence time is
needed to allow the moisture to diffuse from the centre of the particle, there
is little to be gained in improving the vapour-phase heat and mass transfer
to the outside of the particle. Conversely, there may be little point in using a
highly sophisticated model to calculate the external processes very accu-
rately when they are not the limiting factor for the overall drying.
Figure 2 shows results from a pneumatic conveying dryer (Kemp and
Oakley 1997) which show that that there is an optimum range of gas flows
to maintain product at the specified moisture content (0.03 kg/kg or less).
At low gas velocities the dryer is heat and mass balance limited; the mass
flow of air WG is so low that the heat content of the air is insufficient to
supply the latent heat of evaporation required to remove the moisture.
At high gas velocities the kinetics become the limiting factor; the solids
residence time in the duct becomes insufficient to allow the material to dry
adequately. The simulation was generated using an incremental model,
fitted to existing plant data, and also showed the effect of an increase in
throughput.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1709


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 2. Plot of outlet moisture content against gas velocity and solids throughput
for pneumatic conveying dryer.

SCOPING DESIGN CALCULATIONS

In scoping calculations, some approximate dryer dimensions are


obtained based mainly on a heat and mass balance, without measuring a
drying curve or other experimental drying data. They allow the cross-
sectional area of convective dryers and the volume of batch dryers to be
estimated surprisingly accurately, but are less effective for other calculations
and can give highly over-optimistic results.
When designing a continuous dryer, the required solids throughput WS
and the inlet and outlet moisture content XI and XO are known, as is the
ambient humidity YI. For convective dryers, if the inlet gas temperature TGI is
chosen, the outlet gas conditions (temperature TGO and humidity YO) can
be found, either by calculation or (more simply and quickly) by using the
constant-enthalpy lines on a psychrometric chart. It may be necessary to
allow for heat losses, sensible heating of solids and the heat of wetting,
especially if tightly bound moisture is being removed. The gas mass flowrate
WG can now be calculated as it is the only unknown in the mass balance on the
solvent:
WG ðYO  YI Þ ¼ Wev ¼ WS ðXI  XO Þ ð3Þ
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1710 KEMP AND OAKLEY

A suitable gas velocity UG along the dryer is now chosen, typically


20 m/s for a flash dryer, 0.5 m/s for a fluidised bed, 3 m/s for a cocurrent
rotary dryer. For through-circulation and dispersion dryers, the cross-
sectional area A is given by WG ¼ GIUGA, and the diameter can then be
calculated. The result is usually accurate within 10%, and can be further
improved by better estimates of velocity and heat losses.
The method gives no information about solids residence time or dryer
length. A minimum drying time can be calculated by evaluating the
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

maximum (unhindered) drying rate Ncr, assuming gas-phase heat transfer


control and free moisture. Alternatively, it may be assumed that first-order
kinetics apply throughout the drying process. These crude methods are not
recommended, and can give serious underestimates of the required drying
time, especially if unhindered drying is assumed. It is much better to measure
the drying time experimentally and apply scaling methods.
For continuous contact dryers, the vital factor is the area of the heat
transfer surface, AS. This can be found from the equation:

Wev ev
AS ¼ ð4Þ
hWS TWS

The latent heat of evaporation lev should allow for bound moisture
and heating of solids and vapour to the final temperature. A typical wall-to-
solids heat transfer coefficient hSW for the given dryer type should be used.
The calculation is less accurate than the one for convective dryers.
It implicitly assumes that the heat transfer rate is the overall limiting factor.
If the drying process is strongly limited by falling-rate drying kinetics, the
calculated size of dryer corresponding to the given heating surface AS may
not give sufficient solids residence time to reach the desired final moisture
content. Again, experimental measurement of a drying curve is strongly
recommended.
For batch dryers where the batch size is stipulated, the problem
becomes very different, as the requirement is simply that the dryer can physi-
cally contain the volume of the solids, and the dryer volume and dimensions
can thus be calculated directly. The difficulty is now to calculate solids resi-
dence time. Equations (3) or (4) can be reversed, but again the calculation will
be inaccurate for falling-rate drying and it is preferable to measure a drying
curve and use a scaling calculation, as outlined in the next section. However, it
is possible to compare the surface-area-to-volume ratio of various types of
dryer and deduce how their drying times will compare with each other.
Table 2 gives calculated dimensions for various batch contact dryers
of volume 20 m3 operating at 50% volumetric fill (i.e. a batch of 10 m3).
The active area is the area of heat transfer surface actually in contact with
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1711

Table 2. Comparative Dimensions and Drying Times for Various Batch Contact
Dryers

Typical Total Active ‘‘UA


h.t.c. Typical D L Area Area Value’’ tS
Dryer Type kW/m2 K L/D m m m2 m2 KJ/s K h

Tumbler/double-cone 0.1 1.5 3.71 5.56 38.91 19.45 1.95 2.78


Vertical pan 0.1 0.5 3.71 1.85 32.37 21.58 2.16 2.51
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Spherical 0.1 1 3.37 3.37 35.63 17.82 1.78 3.04


Filter-dryer 0.1 0.5 3.71 1.85 21.58 10.79 1.08 5.01
Conical agitated 0.1 1.5 3.71 5.56 34.12 21.50 2.15 2.52
Horizontal agitated 0.08 5 1.72 8.60 46.50 23.25 1.86 2.91
HA, agitator heated 0.08 5 1.72 8.60 278.99 139.49 11.16 0.48

the solids. Assuming unhindered drying with a typical heat transfer coefficient
for each dryer type, the drying time for a given batch has been calculated.
The table shows that drying times are comparable for most types of
double-cone (rotating vacuum batch) and vertical agitated dryer, and for
horizontal agitated dryers with only the jacket heated. The drying time for
a simple filter-dryer is longer, as the bottom filter plate cannot be heated.
It can also be shown that drying times for nearly all contact dryers increase
approximately proportionately with diameter and with the cube root of
batch size. Much larger heating surfaces in a given volume, and hence
shorter drying times, are obtained by heating the internal agitator of a
horizontal agitated (paddle) dryer, and this will be the preferred contact
dryer type for large batches. This of course assumes that the rate of heat
supply is the limiting factor for drying. If a minimum residence time is
required to allow removal of tightly bound moisture, there will be little or
no gain from providing very large amounts of heat transfer area.

MODELS INCORPORATING EXPERIMENTAL DATA

Heat and mass balances and scoping design calculations can be


performed using only tabulated data from literature, and are confined to an
equipment model. We now cover models which incorporate the material
model as well, requiring experimental data from drying kinetics tests.
There are three basic types:
. Integral or scaling models. These treat the dryer as a complete
unit, and are generally based on scaling up a batch drying curve
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1712 KEMP AND OAKLEY

recorded by experiment. These methods are used for layer dryers


(tray, oven, horizontal-flow band and vertical-flow plate types)
and for a simple estimate of fluidised bed dryer performance.
As in scoping design, the air velocity is again simply averaged over
the dryer, or over a given section or region of it.
. Incremental models. These track the local conditions of the gas
and particles through the dryer, mainly in one dimension. They are
especially suitable for cocurrent and countercurrent dryers, e.g.,
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

flash (pneumatic conveying) and rotary dryers. The air conditions


are usually treated as uniform across the cross-section, and depend-
ent only on axial position. This method can also be used to deter-
mine local conditions (e.g. temperature) where a simpler model has
been used to find overall drying rate.
. Complex three-dimensional models, e.g., CFD (computational
fluid dynamics), aiming to solve the gas conditions and particle
motion throughout the dryer. They are the only effective models
for spray dryers because of the complex swirling flow pattern; they
can also be used to find localised conditions in other dryers.
The integral model is a lumped parameter model, whereas the two
types of detailed design models are distributed parameter models.

Integral Model

The integral model was first suggested for fluidised beds by Vanecek
et al. (1964, 1966). The mean outlet moisture content is given by summing
the product of the particle moisture content and the probability that it
emerges at time t:
Z1

X¼ EðtÞXðtÞ dt ð5Þ
0

Implicit assumptions are therefore:


(a) the drying curve function XðtÞ must be experimentally measured;
(b) it must be possible to scale the curve to the new operating
conditions;
(c) the results must be capable of scale-up from small-scale test to
large-scale dryers;
(d) the residence time function EðtÞ must be known.
Hence for (a), accurate small-scale drying kinetics tests will be required,
and for (b) and (c) a calculation method for scaling to new operating
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1713

conditions is needed. The most useful concept here is the characteristic drying
curve, originally introduced by van Meel in 1958. The specific drying rate
concept (Moyers, 1994), described in more detail under layer dryers, is also
useful. For batch and pure plug-flow dryers, E(t) is a delta (spike) function;
all particles have the same residence time, so only one point on the drying
curve need be considered.
Scaling calculations can be remarkably simple, especially for straight-
forward layer dryers (tray, band, vacuum agitated, rotating double-cone,
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

etc.). For example Oakley (2001) has outlined a number of basic scale-up
criteria which are simple ratios of heat transfer coefficients, solids mass, cross-
sectional area and dryer diameter. These rules work well—as long as the set of
criteria needed to satisfy the rules are met. It is therefore vital to understand
the assumptions involved for each type of scale-up equation.
The principle of scale-up calculations is in essence the same as the
integral model, i.e., measure a batch drying curve (or, at least, inlet and
outlet moisture content) for a given residence time, then scale to new
conditions or equipment. The equations are similar to those for a scoping
performance calculation, but allow for the falling rate period correctly.
Hence scaling calculations can be used to improve the result for a scoping
calculation.
Most dryer manufacturers have traditionally relied on performing
pilot-plant tests and scaling the results to a new set of conditions on a
dryer with greater airflow or surface area. It can be seen that in practice,
they have therefore been using the integral model, although not necessarily
recognising it as such. Moreover, the empirical rules employed have generally
been based on the external driving forces (temperature, vapour pressure or
humidity driving forces). By implication, therefore, a characteristic drying
curve concept is being used, scaling the external heat and mass transfer and
assuming that the internal mass transfer changes in proportion.

Incremental Model

The one-dimensional incremental model is a key analysis tool for sev-


eral types of dryer. A set of simultaneous equations is solved at a given
location and the simulation moves along the dryer axis in a series of steps
or increments—hence the name. The procedure may be attempted by hand if
a few large steps (say 5–10) are used, but for an accurate simulation, a com-
puter program is needed and thousands of increments may be used (Fig. 3).
Increments may be stated in terms of time (dt), length (dz) or moisture
content (dX ). A set of six simultaneous equations is then solved, and
ancillary calculations are also required, e.g., to give local values of gas and
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1714 KEMP AND OAKLEY


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 3. Principle of the incremental model.

solids properties. The generic set of equations (for a time increment t) is
as follows:
Heat transfer to particle: qQp ¼ hPG AP ðTG  TS Þ ð6Þ
dX
Mass transfer from particle: ¼ functionðX; Y; TP ; TG ; hPG ; AP Þ ð7Þ
dt
dX
Mass balance on moisture: WG Y ¼ ðWS XÞ ¼ WS t ð8Þ
dt
Q t  ev mP X
Heat balance on particle: TS ¼ P ð9Þ
mP ðCPS þ CPL XÞ
Heat balance for increment:
W ðC þ CPL XÞTS þ WG ð0 þ CPY TG ÞY þ QWt
TG ¼ S PS ð10Þ
WG ðCPG þ CPY YÞ

Particle transport: z ¼ US t ð11Þ

The mass and heat balance equations are the same for any type of
dryer, but the particle transport equation is completely different, and the
heat and mass transfer correlations are also somewhat different as they
depend on the environment of the particle in the gas (i.e., single isolated
particles, agglomerates, clusters, layers, fluidised beds or packed beds). The
mass transfer rate from the particle is regulated by the drying kinetics, and is
thus obviously material-dependent (at least in falling-rate drying).
The model is effective and appropriate for dryers where both solids
and gas are approximately in axial plug-flow, such as pneumatic conveying
and cascading rotary dryers. However, it runs into difficulties where there is
recirculation or radial flow.
The incremental model is also useful for measuring variations in local
conditions such as temperature, solids moisture content and humidity along
the axis of a dryer (e.g. plug-flow fluidised bed), through a vertical layer
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1715

(e.g. tray or band dryers) or during a batch drying cycle (using time incre-
ments, not length). It can be applied in these situations even though the
integral model has been used to determine the overall kinetics and drying time.

Computational Fluid Dynamics (CFD)

CFD provides a very detailed and accurate model of the gas phase,
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

including three-dimensional effects and swirl. Where localised flow patterns


have a major effect on the overall performance of a dryer and the particle
history, CFD can give immense improvements in modelling and in under-
standing of physical phenomena. Conversely, where the system is well-mixed
or drying is dominated by falling-rate kinetics and local conditions are
unimportant, CFD modelling will give little or no advantage over conven-
tional methods, but will incur a vastly greater cost in computing time.
The classic application of CFD in drying is for spray dryers, but it has
also been useful for other local three-dimensional swirling flows, e.g.,
around the feedpoint of pneumatic conveying dryers (see below), and for
other cases where airflows affect drying significantly, e.g., local overdrying
and warping in timber stacks (Langrish, 1996). A recent meeting in France
(AFSIA 2001) covered a number of other cases where CFD has aided
modelling of airflows in dryers, including fluidised beds, superheated
steam and infra-red dryers, for materials including pharmaceuticals, clays
and ceramics, Emmental cheeses and smoked sausages.

DRYING KINETICS

Theoretical Models

If the falling-rate period is to be modelled effectively, it is vital to choose


an appropriate drying kinetics model; in particular, whether a distributed-
parameter or lumped-parameter model should be used. Distributed param-
eter models try to use fundamental physical equations and quantities. In
drying, a large number of non-linear simultaneous equations will be required.
This approach can be successful, as long as all the required parameters can be
measured for the given material within reasonable limits of error. However, if
some of the quantities are difficult to measure, e.g., pore size distribution,
pore tortuosity and diffusion coefficients in solids, the theory will not yield
usable numerical results. In drying kinetics, the distributed parameter
approach is exemplified by the work of Luikov (1935, 1966) and Whitaker
(1977, 1980), among others. However, there are only a handful of materials,
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1716 KEMP AND OAKLEY

out of the millions of solid products, for which all the parameters required in
the equations have been measured.
Moreover, some distributed-parameter models tend to be divergent
and are very sensitive to errors in the basic parameters. For example,
Kemp and Oakley (1997), for their incremental model for pneumatic
conveying (flash) dryers, found that a 10% error in Sauter mean particle
size dp(SM) had a substantial effect on the predicted drying rates. In a real
drying situation with a wide range of particle sizes and shapes in the feed,
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

the error in measuring dp(SM) can easily exceed 10%. Hence, even for an
apparently fundamental quantity such as particle size, the results may need
to be back-fitted against experiment.
Lumped parameter models require a smaller number of equations, and
the material is characterised by a few ‘‘lumped’’ parameters. These typically
combine aspects of several different physical phenomena; for example, the
various competing moisture transport processes in a solid may be modelled
using a pseudo-diffusion coefficient, treating the system as if it were control-
led only by gas- or liquid-phase diffusion. Obviously, lumped parameters
cannot be theoretically predicted and must be measured, and it is dangerous
to extrapolate them to widely different operating conditions. However, if the
physical significance and the limits of applicability of lumped parameters are
clearly understood, interpolation and limited extrapolation to new conditions
are often possible.
The characteristic drying curve is a classic example of a lumped
parameter model. Indeed, one can regard a drying curve as an extreme
example of a lumped parameter, expressing the relationship between moist-
ure content and drying time.
Obviously the major question is whether lumped parameters, whose
theoretical basis is limited, are adequate to describe real drying processes.
Fyhr and Kemp (1998b) simulated a drying system accurately by an
advanced model and approximately by the characteristic drying curve.
The responses of the two models to changes in operating conditions were
remarkably similar. It was shown that the CDC introduced three main
errors due to the simplification of the theory, but in nearly all circumstances,
these errors acted in opposite directions and tended to cancel each other out.
Furthermore, they used a simple power-law characteristic drying curve,
fitted in two sections, to approximate a complex rate-moisture curve obtained
from an advanced model, and then back-calculated the moisture-time curve.
The results are shown in Fig. 4. On the Krischer chart, the CDC does not look
a particularly close approximation, but when this is converted to a drying
curve, the approximation is indistinguishable from the original curve over
most of the range; if a drying time were read from either curve, the values
would be within 2% of each other. However, at the bottom end, the curves
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1717


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 4. Effect of an approximate data fit on rate–moisture and moisture–time


curves.

diverge sharply; the time to dry to X ¼ 0.02 would be read as 110 s from the
characteristic curve but is 200 s in reality, which would lead to a seriously
undersized dryer. However, this is not because the characteristic curve model
is faulty, but because the fit to the data at the bottom end was poor. If the data
below  ¼ 0.1 is refitted with a further section of power–law curve, the CDC
result again becomes very close to the results from the advanced model.
The significance of this is that it is the moisture–time curve which, by implica-
tion, is used to determine the required drying time and to size the dryer.
These results show that the simplified CDC model, with a reasonable fit to
experimental data, gives an accurate determination of the drying time and is
hence suitable for practical design of industrial dryers.

Experimental Measurement Techniques

A good review of methods for measuring drying kinetics of particles


was given by Keey (1992), and recommendations for best practice for
processing the data obtained were given by Kemp et al. (2001).
An important question of principle is whether one should seek to
measure the fundamental drying kinetics of a single particle, or that of an
assemblage of particles. In the latter case, layer effects will distort the drying
kinetics. A pilot-plant test on a similar dryer to the full-scale unit will
minimize scale-up and extrapolation, but the results are specific to that
dryer type. Tsotsas has given a simple but elegant representation of the
principles involved (see Fig. 5).
Experiments can be made on a variety of scales:
– Single particle measurements, using a microbalance or drying tunnel.
These give the fundamental single-particle drying kinetics, but for
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1718 KEMP AND OAKLEY


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 5. Tsotsas’ representation of alternative drying kinetics test methods.

small, light particles it is difficult to get an adequate signal-to-noise


ratio.
– Thin-layer tests, passing gas through a solids layer of about 5 mm.
Gives good experimental curves, but a scale-down calculation
is needed as the kinetics are distorted by the layer thickness,
especially for particles less than 0.5 mm diameter.
– Small-scale tests on miniature versions of the dryer, e.g., single
trays, small fluidised beds. Drying kinetics from these tests cannot
safely be extrapolated to other types of dryer.
– Pilot-plant tests; again, the results will be dryer-specific.
For distributed-parameter and detailed design models, such as the
incremental model, the fundamental drying kinetics are required and one
of the first two methods should be used. In contrast, scaling methods are
lumped-parameter methods which rely on the kinetics measurements being
made on similar apparatus to the full-scale dryer, so small-scale or pilot-
plant tests must be used to obtain the drying curves in these cases.
SPS have a convective drying kinetics rig incorporating a thin-layer
test module and a fluidised bed/deep bed module (Figs. 6 and 7). Air (or
nitrogen) is conditioned to a given flowrate, humidity and temperature and
supplied to the required module. The humidity of the inlet and exhaust air is
continuously measured using an infra-red gas analyser (IRGA), and
the change in humidity is directly proportional to the drying rate, which
is then integrated to give a drying curve. The modules are shown below. The
thin-layer apparatus was developed during joint projects with the University
of Canterbury, New Zealand (Langrish et al., 1991a, 1991b). A recent
survey of the underlying theory was made by Fyhr and Kemp (1998a).
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1719


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 6. Gas conditioning system of the SPS drying kinetics rig.

Figure 7. Fluidised bed and thin layer drying modules.

Hirschmann et al. (1998) compared the drying curves obtained from


both single-particle and thin-layer kinetics experiments, showing that they
were approximately equivalent. However, the single-particle experiments
had to be performed at much lower relative velocities than the thin-layer
tests.
Keech et al. (1996, 1998) describe the use of a microbalance to measure
drying kinetics at very low moisture contents and under high vacuum.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1720 KEMP AND OAKLEY

APPROPRIATE MODELS FOR DISPERSION DRYERS

Having described the main types of dryer model, we now show how
they are applied to the four main types of dispersion dryer—fluidised beds,
pneumatic conveying (flash) dryers, cascading (direct) rotary dryers and
spray dryers.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Fluidised Bed Dryers

The integral model is used to determine drying kinetics and solids


residence times, although the incremental model is also useful to determine
local temperatures during a batch, at different distances along a plug-flow
continuous dryer and at different heights in the bed. Drying kinetics can be
measured in a small-scale fluidised bed.
The basic method for thermal design of fluidised bed dryers was given
by Reay and Allen (1982a,b) and extended by Reay (1989), McKenzie and
Bahu (1991) and Bahu (1994). The drying kinetics are measured in a pilot-
scale fluidised bed and the time axis of the drying curve is scaled for a given
change in moisture content X using these equations:
2 ðmB =AÞ2 G1 ðTGI  Twb Þ1
Type A (fast drying material): Z¼ ¼ ð12Þ
1 ðmB =AÞ1 G2 ðTGI  Twb Þ2
2 ðTGI  Twb Þ1
Type B (slow drying material): Z¼ ¼ ð13Þ
1 ðTGI  Twb Þ2
For Type A materials, drying rate is accelerated by higher gas
velocities and higher inlet gas temperatures, but reduced by increasing bed
depth. For Type B materials, drying rate is independent of gas velocity and
bed depth but was still affected by gas temperature. The vast majority of
materials obey Type A rules, even in falling-rate drying; for example, iron
ore, silica gel, ion exchange resin. The only material found to exhibit Type B
drying throughout was wheat, where drying is limited by the very slow
diffusion through the cell wall.
In Reay and Allen’s original papers, well-mixed fluid bed dryers were
considered and the temperature driving force term (TGI  Twb) was replaced
by a vapour pressure term ( pYs  pYI), where pYs is the saturation vapour
pressure at the bed temperature. However, for plug-flow and batch dryers,
where bed temperature varies continuously, the form in Eqs. (12) and (13) is
more convenient.
The normalisation factor Z allows the drying curve function to be
scaled from the initial test conditions to the chosen gas velocity, bed
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1721

depth and temperature of the full-scale dryer. Gas mass velocity G is used
instead of gas velocity UG as it is independent of temperature. Likewise, bed
weight per unit area (mB/A) is preferred to bed depth z, as the former does
not change as the bed fluidises, expands and repacks.
The SPS design method has been successfully used over the last 20
years to design and debottleneck many industrial dryers. However, it has
some weaknesses:
a. the equations were derived from experiment and the theoretical
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

basis was not fully explained;


b. the drying kinetics is almost invariably falling-rate throughout,
implying internal mass transfer control, yet the equations predict
the main dependence to be on external conditions (temperature,
gas velocity and bed depth) and use a heat transfer driving force,
implying unhindered drying;
c. the results were not fully reliable for large changes in temperature;
d. ways had to be found to apply the equations to beds where addi-
tional heat is added by means of internal heating coils or plates;
e. some materials show a transition from Type A to B normalisation
behaviour in the later stages of drying.
McKenzie and Bahu (1991) showed that with internal heating coils,
the normalisation factor was proportional to the total heat input into the
bed; it did not matter how the heat was distributed between the hot air and
the coils.
More recent analysis within SPS confirms the theoretical basis of the
equations above. Because heat and mass transfer from gas to particles in
fluidised beds is excellent, it has long been known that the majority of the
heat transfer takes place in a narrow layer close to the distributor. Indeed,
the Type A scaling function for bed depth implicitly assumes this (increasing
the bed depth simply increases the proportion of solid in the ‘‘dead zone’’
where no significant drying occurs, and reduces the frequency at which
particles recirculate vertically into the active zone). In effect, thermal equilib
rium between gas and particles is reached rapidly, and the number of trans-
fer units (NTU) approaches infinity. So although the drying kinetics of
individual particles are hindered (falling rate), the overall process is heat
and mass balance controlled, because of the deep bed (in heat and mass
transfer terms) and the very high NTUs. This explains the paradox in
(b) above. Since control is by total heat input, not heat transfer, the scaling
factor in the equations should really be (HGI  HGO), where H is the dry gas
enthalpy, or (TGI  TGO), not (TGI  Twb). This is computationally incon-
venient because TGO at any point in the cycle is equal to solids
temperature TS, and this varies throughout the cycle. However, where
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1722 KEMP AND OAKLEY

significant drying is taking place with high NTU’s, the outlet gas reaches the
adiabatic saturation temperature Tas. Not only is this easy to calculate, but
it explains why the (TGI  Twb) scaling factor has been successful, as the wet
bulb and adiabatic saturation temperatures are coincidentally almost equal
for the air–water system.
Now the equation for NTU’s through a bed of height z is:
Z YO Z Zz
dY f kY a z
¼ dz ¼ f  NTU dz ð14Þ
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

YY ðYwb  YÞ G 0 0
ðYwb  YO Þ
¼ expðf  NTU  zÞ ð15Þ
ðYwb  YI Þ
This assumes that all the particles have had similar drying times and
have the same value of f, i.e., a batch bed or a small region of a plug-flow
continuous unit. Since the evaporative flux per unit area of the bed Jev ¼
G(YO  YI), a little manipulation gives:
Jev;2  
¼ ðYwb  YI Þ 1  ef NTUz ð16Þ
G
For scaling drying curves between two sets of conditions, this equation
becomes:

Jev;2 fGðYwb  YI Þ1 ð1  ef NTU:z Þg1


¼ ð17Þ
Jev;1 fGðYwb  YI Þ2 ð1  ef NTU:z Þg2

We can now substitute for the evaporative flux per unit area:
Wev dX mB
Jev ¼ ¼ ð18Þ
AB dt AB

Over a given increment of moisture content, and using a temperature


driving force instead of a humidity driving force, this gives:
2 X1 ðmB =AÞ2 G1 ðTGI  Twb Þ1 ð1  ef NTUz Þ1 2
¼ ¼ ¼Z ð19Þ
1 X2 ðmB =AÞ1 G2 ðTGI  Twb Þ2 ð1  ef NTUz Þ2 1
since we use the same moisture content increment X on both the original
and scaled curves. Thus we have a new formula for the normalisation factor
Z. Its similarity to the two existing formulae is clear. Moreover, for fast
drying materials where the NTU is high, the exponential terms tend to zero
and the formula reduces to Eq. (12).
Conversely, for slow drying materials, where f is low and the gas
emerging from the bed is not saturated, the exponential term becomes
significant and can be expanded to f kYaz/G. Assuming that f, kY,  and a
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1723

are the same for both cases, and noting that bed depth z is proportional
to mB/AB, we end up with:
2 ðmB =AÞ2 G1 ðTGI  Twb Þ1 fðmB =AÞ1 =G1 g ðTGI  Twb Þ1
Z¼ ¼ ¼ ð20Þ
1 ðmB =AÞ1 G2 ðTGI  Twb Þ2 fðmB =AÞ2 =G2 g ðTGI  Twb Þ2
which is the Type B formula, Eq. (13). Thus, for a bed section with a small
but finite number of modified transfer units ( f  NTU), the rigorous formula
in Eq. (19) can be used; elsewhere, the simplified Type A and Type B
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

formulae will apply.


Some workers have queried the use of the integral model in fluidised
beds, suggesting that such a basic model cannot be adequate for modelling
drying behaviour and that more sophisticated models will be required,
including bubble growth and local gas and particle flow patterns, or even
using CFD to track the history of the individual drying particles. However,
because the drying process is largely heat and mass balance limited, the
simple model is adequate in virtually all circumstances. Better modelling of
the internal moisture transport within the particle could help. The localised
temperatures of particles during the batch or along the bed, and hence the
corresponding gas temperature, cannot be predicted by a simple integral
model. An incremental model can be used to give this additional informa-
tion, but it is not necessary for drying rates, even where a split distributor
(plug flow continuous) or varying airflows (batch dryers) are used, as the
different portions of the drying curve can be scaled separately. Variations in
solids residence time and outlet moisture content in a non-ideal plug-flow
dryer can be predicted by backmixing calculations (Reay, 1989, Fyhr, Kemp
and Wimmerstedt, 1999).
The incremental model can also be used to predict local conditions at
various heights within the fluidised bed, as shown by Fyhr and Kemp (1999).
Figure 8 shows results from the Fyhr and Kemp model with an inlet
air temperature of 110 C. It confirms the rapid thermal equilibration near the
distributor; in the upper layers of the bed, the gas and solids temperatures
are identical throughout the drying process.
An alternative approach is to use a distributed parameter model,
measure single particle kinetics and scale up to the full size bed using theory.
This approach is described by Tsotsas (1994) and Burgschweiger et al. (1999).

Pneumatic Conveying Dryers

The cross-sectional area is found from the scoping design calculation.


To find length, the one-dimensional incremental model is used, tracking
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1724 KEMP AND OAKLEY


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 8. Predicted temperature profiles through a fluidised bed.

vertically along the duct. For this type of dryer, particle motion, heat transfer
and drying kinetics require specific treatment. The model is summarised by
Kemp and Oakley (1997). Drying kinetics are measured by a thin-layer
method or by suspending single particles in a drying tunnel.
For most purposes, the gas-to-particle heat transfer coefficient may be
obtained by the Ranz and Marshall (1952) equation. However, for the high
relative gas velocities found near the feedpoint of pneumatic conveying
dryers (and the entry zone of spray dryers), a turbulent boundary layer may
develop and SPS have found that a modified equation developed from a
suggestion by Weber (see Kemp, Bahu and Pasley, 1994) gives better results
for Re>10:

Nu ¼ 2 þ 0:5Re0:5 Pr0:5 þ 0:05Re0:8 Pr0:33 ð21Þ

The first term corresponds to the heat transfer to a sphere in undis-


turbed air, the second is the heat transfer through the laminar boundary
layer and wake, and the third term accounts for the turbulent boundary
layer. The equations give similar results at low Reynolds numbers, but
at Re ¼ 1000 (corresponding to a 1 mm particle entering a gas stream
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1725

at 20 m/s in a pneumatic conveying dryer, or a 200 mm droplet thrown from


an atomiser at 100 m/s) the Modified Weber correlation predicts a heat
transfer coefficient 50% higher than Ranz and Marshall.
Kemp and Gardiner (2001) used the incremental model for optimisa-
tion of an existing flash dryer and showed that there was a clear optimum air
velocity. At low air velocities, the inlet air had insufficient heat to perform
the drying duty (heat balance limitation), whereas at high air velocities, the
particles were blown out of the dryer too quickly (kinetics limitation).
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Measuring the drying kinetics presents major practical problems, as


the residence time of the particles in a typical pneumatic conveying dryer is
of the order of one second, and the temperature and flowfield to which the
particle surface is exposed varies widely during this period. In practice, the
only moisture which will be removed will be that at the surface or within a
few microns of it. It has been postulated that experiments on the SPS thin-
layer drying kinetics rig at a number of different residence times from a few
seconds up to a couple of minutes, at various gas temperatures and velo-
cities, could provide the necessary information. Only the initial and final
moisture contents would be measurable, using an oven test or similar.
Results of recent experiments on silica gel by Kemp and Fernandez
(2001) using this technique are shown in Fig. 9. The link between drying
time and moisture content is clear. Comparison with results at the same

Figure 9. Results of short residence time thin-layer drying experiments on silica gel.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1726 KEMP AND OAKLEY

temperature from a pilot-plant flash dryer show that the moisture reduction
in the pilot-plant dryer (residence time approximately one second) is about
10%, which was achieved in about five seconds in the thin-layer test. This
tentatively suggests that a thin-layer experiment can give an indication of
the drying kinetics in a flash dryer. The faster drying in the pilot plant can
be ascribed to the very high relative velocities at the entry of the particles
(initially 15 m/s as against the constant 4 m/s in the thin-layer test), and the
higher number of transfer units in the layer as compared to discrete par-
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

ticles. An interesting result was that the medium (425–1000 micron) particles
dried fastest, the large (1–4 mm) particles dried more slowly, but the fine
particles (below 212 mm) dried most slowly of all. The powder agglomerated
severely and the hot air could not reach the interior, even after a minute or
more of drying.
Flow patterns for particle motion in a pneumatic conveying dryer are
essentially linear along the duct, but CFD was found to be useful for studying
the feedpoint region by Kemp, Oakley and Bahu (1991), where it helped to
explain puzzling observations about particle velocity and recirculation zones.
Figure 10 shows that two unexpected phenomena were observed.
Firstly, the particles were observed to accelerate up the duct and then slow
down slightly before accelerating again to their final steady-state velocity.
This could not be predicted by the standard one-dimensional incremental
model, even when wall friction was taken into account (which gave excellent
predictions of the final steady-state velocity). Secondly, the particles were
observed to ‘‘dip’’ on entering the duct, whereas the one-dimensional force
balance predicted that they would immediately be accelerated upwards.
CFD was able to predict and explain these observations. The simula-
tion predicted the ‘‘dipping phenomenon’’ and showed that it was dependent

Figure 10. Observed particle motion at feedpoint of pneumatic conveying dryer.


MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1727


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 11. Particle motion predicted by CFD at feedpoint of pneumatic conveying


dryer.

on the rate and radial velocity of the solids feed. The momentum transfer to
the large mass of solids entering the duct caused a substantial drop in local gas
velocity and a ‘‘dead zone’’ where little acceleration took place—see Fig. 11.
The particles therefore dropped until they encountered the main gas flow
and were accelerated around the dead zone, with both gas and particles
reaching higher velocities in the constriction. In effect, the gas had created
its own venturi at the feedpoint. CFD was also able to show how the
feedpoint could have been designed better, to avoid the dead zone.
Finally, drying in the cyclone which follows the dryer is a neglected but
important area, as the solids residence time in the cyclone can be several times
greater than that in the main duct. Kemp et al. (1998) studied this in some
depth, using both experimental and theoretical methods. They concluded that
the cyclone could be modelled most effectively by treating it as an additional
section of duct. It is important to know the solids residence time , and
experiments with silica gel with particle sizes in the range 200–1000 mm
showed that  increased strongly with particle density P and cyclone
diameter D and weakly with particle diameter dP, while it was largely
independent of gas velocity or solids flowrate. The following correlation
was a best fit to the results:
 ¼ 0:06dP0:3 P DLU0:3
G ð22Þ
This gives a calculated residence time  of 1.5 s for 300 mm silica gel
particles of density 1200 kg/m3 in a cyclone of diameter D ¼ 350 mm with
inlet gas velocity UG ¼ 15 m/s. This correlation cannot, of course, be
extrapolated safely to different situations; for example, other workers’
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1728 KEMP AND OAKLEY

results suggest that a completely different trend applies for particle dia-
meters below 100 micron. Also it does not allow for changes in geometry,
particularly cone angle. Extensive further measurements on particle resi-
dence times and solids flow patterns in cyclones were given by Saruchera
(1999). However, there is still no full theoretical explanation for the resi-
dence time trends observed experimentally. Kemp et al. (1998) also noted
that CFD simulation had been unsuccessful, because of the large number of
particle–particle interactions, but advances in computing techniques may
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

now make this a possibility.

Cascading Rotary Dryers

The one-dimensional incremental model is used, tracking horizontally


along the drum axis. Particle motion and heat transfer require specific
treatment. The model is summarised by Kemp and Oakley (1997). Drying
kinetics are measured by the thin-layer method.
The cross-sectional area and drum diameter can be found easily from a
scoping calculation, using estimated inlet and outlet gas temperatures and
mean gas velocities (typically 3 m s1 for a cocurrent dryer, 2 m s1 for a
countercurrent unit). As usual, drying time and dryer length cannot be
found accurately by scoping methods. In practice, dryer length has often
been chosen simply by using a standard L/D ratio of between 5 : 1 and 10 : 1
(typically 8 : 1), since the dryer diameter is already known. Any resulting
practical problems (e.g., with insufficient residence time) are overcome by
altering the holdup by means of internal weirs.
To find the required solids residence time  by a real calculation,
Friedman and Marshall (1949) and Saeman and Mitchell (1954) proposed
empirical and semi-theoretical correlations. Surprisingly, these relatively
crude correlations are still reproduced in quite recent papers, although
models based on sounder scientific principles have been in the literature
for many years. Early examples include Schofield and Glikin (1962) and
Kelly (1969a,b and 1977). Baker (1983) showed that these methods were
inaccurate for prediction of industrial dryer performance. This provided an
incentive for the development of an improved two-stream model of particle
motion by Matchett and Baker (1987), which considered the airborne and
dense phases separately. The model was verified on different dryers ranging
from pilot to commercial scale by Matchett and Baker (1988); agreement
between calculated and measured holdup was always within 20%. Cao
and Langrish (1999) demonstrated that residence time predictions from the
Matchett and Baker model were superior to those of Friedman and
Marshall (1949) and Saeman and Mitchell (1954). The Matchett and
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1729


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 12. Airborne-and-dense-phase model for cascading rotary dryers.

Baker two-stream model was combined with an incremental model along


the dryer axis to calculate the length by Langrish (1988), with the drying
kinetics being measured in a thin-layer test.
The original Matchett and Baker model was a ‘‘parallel’’ model, treating
the solids as two streams flowing together through the dryer and exchanging
solids continuously. It has been found computationally more convenient
to use a ‘‘series’’ model, where the particles are tracked through both the
airborne and dense phases, as in Fig. 12, and a simplified form of this model
could be calculated on a spreadsheet (Papadakis et al., 1992). The advantage
is that the calculated residence times in the airborne and dense phases,  G
and  S, are physically realistic, and the dryer length is simply the sum of the
distances travelled in the two phases:
 ¼ G þ S ð23Þ
L ¼ G UP1 þ S UP2 ð24Þ
Typically, particles spend 90–95% of the time in the dense phase.
Virtually all the drying takes place during the airborne phase, and axial
particle motion is very different between the two phases. Particle movement
along the dryer occurs by means of four separate mechanisms, two in the
airborne phase and two in the dense phase (Fig. 13):
(i) Gravitational, due to the slope of the drum.
(ii) Drag of the gas on the airborne particles (negative for counter-
current flow).
(iii) Bouncing of the particles on impact with the bottom of the dryer.
(iv) Rolling of the particles in the bed at the bottom of the dryer,
especially for ‘‘overloaded’’ dryers.
The gravity component (i) is easy to calculate, but the drag force (ii) has
to be obtained from experimental correlations (Baker, 1992) and it is wise to
verify it experimentally if possible, using residence time measurements at one
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1730 KEMP AND OAKLEY

or more suitable air velocities. Combining these components gives UP1.


The bouncing and rolling mechanisms are not susceptible to theoretical
analysis. The dense phase velocity UP2 should therefore be evaluated by
residence time measurements, preferably in the actual dryer, at zero airflow.
For scale-up purposes and changes in operating conditions it is found that the
dense phase velocity can be characterised in terms of a dimensionless dense
phase velocity number, a, given by the equation:
UP2
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

a¼ ð25Þ
ND tan
The results can thus be scaled to other values of rotational speed N,
drum diameter D and slope .
The remaining problem is to predict the gas and solids phase residence
times,  G and  S. We make the assumption that all drying takes place in the
airborne phase, with the dense phase simply being for ‘‘recovery’’ (redistri-
bution of internal moisture from the centre of the particle to the surface).  G
is then calculated from standard vapour-phase heat and mass transfer equa-
tions, allowing appropriately for drying kinetics. For heat transfer, Kemp
and Oakley (1997) suggested using Hirosue’s modification of the Ranz and
Marshall equation for isolated particles, but for most particles Langrish
et al. (1988) found that the Kothari correlation for fluidised beds, as
reported by Kunii and Levenspiel (1991), is more appropriate.
Conversely,  S is found by relating it to  G by geometry, calculating
the time of flight from the top of the dryer to the bottom tf and the time
required for the particle to be lifted by the flights td. The ratio between them
will be  G/ S. For a rotation rate of N revolutions per second and an effec-
tive diameter (internal diameter between lips of flights) of De, if the solids
are carried in the flights for an angle 2
before falling:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2De sin

tf ¼ ð26Þ
g
2

td ¼ ð27Þ
2 N
rffiffiffiffiffiffi
 S td
g
¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ð28Þ
G tf N 2 sin
De
This form of the equation would be inconvenient for design because De
is unknown until a decision has been made on the type and geometry of
flights. However, Eq. (28) can be rewritten as:
rffiffiffiffi sffiffiffiffiffiffi
S Kfl g
D
¼ where Kfl ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ð29Þ
G N D 2 sin
De
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1731


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 13. Mechanisms of particle movement along the drum.

Figure 14. Effect of flight width on the relative time factor Kfl.

Here the unknowns have all been rolled into a single dimensionless
parameter Kfl . In the simplest form of the model, where it is assumed that all
the particles fall from
pffiffithe
ffi pflights as they reach the top of the dryer,
¼ /2
ffiffiffiffiffiffiffiffiffiffiffiffi
(90 ) and Kfl ¼ ð1=2 2Þ D=De . Since De normally lies between 0.8D (for
small dryers with large flights) and D (for large dryers or small flights), this
gives Kfl ¼ 0.37 0.03.
Figure 14 shows the variation of Kfl with
, with effective diameter De
ranging from 0.8D to 1.0D, for a 1 m diameter dryer rotating at 0.1 rev/s.
Assumptions are:
– particles fall with zero velocity from the outer lip on the flight
– particles are immediately re-entrained into a flight on meeting it
– air resistance on the falling particle can be neglected.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1732 KEMP AND OAKLEY


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 15. Effect of other phenomena on the relative time factor Kfl .

It can be seen that Kfl increases gradually with


, i.e., the increased
time spent being lifted in the flights outweighs the increased time of flight in
the airborne phase up to
¼ 90 . The majority of the solids tend to leave the
flights in the range between
¼ 30 and
¼ 120 , especially for underloaded
drums, so it can be seen that a reasonable mean value for Kfl is about 0.35.
Flight width has only a weak effect in this simple model.
Figure 15 shows the effect of various refinements to the simple model:
(a) if the particles have a significant initial velocity on leaving the
flights, reducing tf;
(b) if particles spend some time in a rolling bed at the bottom of the
dryer before being captured by the flights, increasing td (a typical
situation for an overloaded drum);
(c) if the particles fall to the bottom of the drum before being captured
by the flights, rather than entering the flights part-way through
the revolution.
All of these tend to increase Kfl , but the increase is moderate. Only if
there is a very extensive rolling bed in the bottom of the dryer, such that the
solids residence time there exceeded the time spent being lifted in the flights,
would the mean value of Kfl be expected to significantly exceed 0.5.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1733

Rigorous theoretical models for particle motion from flights have been
developed by Baker (1988), Langrish (1988) and Matchett and Sheikh (1990).
These include the effects of particles falling from the flights at different levels
(affecting both time of fall and time spent in the dense phase), flight geometry,
initial rolling velocity and even Coriolis forces. As with other distributed
parameter models, these models require a large amount of experimental
data to be collected, and the cumulative error can be significant. Thus,
although clearly superior on a theoretical basis, the advanced models often
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

show no definite advantage in industrial practice, especially as the dense


phase solids velocity still has to be found by back-fitting from experimental
results for residence time.
Hence, in the early stages of design, adequate results may be obtained
from this simplified model, using a value of Kfl between 0.35 and 0.5. An
advantage compared with other published models is that the holdup is not
required and there is no need to specify the geometry of the flights. The
model will generally give good predictions of what will happen if the dryer
speed, slope or gas velocity are changed. Note that the factor Kfl has a clear
and simple physical significance. At a later stage, the number and geometry
of flights may be chosen to ensure that the drum is underloaded or design-
loaded with the given solids residence time, and advanced models may be
used to refine the estimate of Kfl . For overloaded drums, however, it is
strongly advisable to calculate the exact holdup in the flights and in the
rolling bed at the bottom at an early stage of the calculation. Moreover,
particles are much more mobile in the axial direction in the rolling bed than
when held in flights, so the dense phase velocity number a would be expected
to increase with the proportion of solids in the rolling bed.
The fact that there is as yet no consensus on how to model rotary dryers
is borne out by the appearance in the early part of 2000 of no less than three
separate papers on this subject, and in particular particle movement. The most
thorough overall review of dryer models was by Cao and Langrish. Other
models were given by Renaud, Thibault and Trusiak, and Shahhosseini,
Cameron and Wang. Both of these covered particle transport only, as does
an even more recent paper by Langrish, Papadakis and Baker (2001).
Practical problems include entrainment of fines, internal particle flow
patterns, and stickiness. Pilot-scale experiments have proved inadequate for
determining some features, particularly entrainment (Langrish, 1992).
Firstly, in small diameter dryers the cascades do not have so much time
to spread out, so the air flow through the cascades is markedly less and no
elutriation occurred. A thin radial ‘‘slice’’ with the same diameter as the full-
scale dryer was therefore constructed, but the curtains were blown straight
out of the dryer axially—the long curtains in a full-scale dryer have a high
inertia and effectively protect the downstream solids.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1734 KEMP AND OAKLEY

An interesting combination of experience and theory to give unexpected


results was reported in a recent paper by Kemp and Milborne (2000).
Flat (180 ) flights were used in the early part of a fertiliser dryer because
of the sticky nature of the feed. However, replacement by deep 100 flights
confined the stickiness problem to a shorter length of the dryer, contrary to
conventional wisdom. The higher capacity of these flights increased the
residence time of the solids in the entry region, allowing them to dry out.
The low capacity flights merely reduced the extent of the falling curtains and
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

shifted the stickiness problem further down the dryer.

Spray Drying

Spray drying can be considered to be the most difficult challenge of all


in dryer modelling, for a number of reasons:
(i) The physical phenomena associated with droplet drying (depen-
dent on particle Reynolds number) do not scale up proportion-
ately with the dryer diameter, so that dynamic similarity cannot
be retained between pilot-plant and full-scale tests.
(ii) There are complex, highly swirling three-dimensional air flows,
and the droplets usually follow a very different trajectory to the
gas streamlines.
(iii) Particles are formed from a liquid feed within the dryer and the
product morphology and quality, as well as the kinetics, are
strongly affected by temperature history.
(iv) It is difficult to conduct drying kinetics tests with liquid feeds,
especially for small droplets.
Factor (i) makes it difficult to scale up reliably from pilot-scale tests,
and has thus inhibited novelty in dryer design. Factors (ii) and (iii) mean
that CFD is needed to gain a fully satisfactory representation of the drying
process for each individual particle; a one-dimensional incremental model is
inadequate. Factor (iv) makes it difficult to obtain the drying kinetics
parameters to feed into the CFD (or any other) model.
Traditional spray dryer design procedures were based on manufac-
turers’ experience with a small number of well-tried chamber configurations.
Scale-up was based on empirical rules and a ‘‘chamber coefficient’’ e. This
was the ratio of the mean gas residence time (which could be calculated
easily) to the mean solids residence time (which could be measured with
difficulty with tracers). This approach took no account of the air flow
patterns inside the dryer. When the chamber coefficient had been obtained
by experiment for a spray dryer of given geometry, it was assumed to apply
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1735

to geometrically similar spray dryers of similar size handling particles of


similar particle size and bulk density. Pilot-plant testing was essential as a
backup, but had to be performed on a large scale because, on small units,
the particles impinge on the walls. If the atomising velocity is reduced to
avoid this problem, the morphology of the particles and their drying behavior
change markedly. Hence, for obvious reasons, manufacturers tried to use
standard dryer geometries as far as possible. New, innovative geometries
would be risky, and would necessitate further extensive trials.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

It was not until the development of modern computing, giving the


power to solve the Navier-Stokes equations in a reasonable time, that
spray dryers could be modelled truly effectively. CFD was first applied to
spray dryers by Crowe (1980). The basic features of his approach have been
followed by most workers since then. A large number of individual droplets,
of various sizes, are tracked through the gas inside the chamber. The local
gas temperatures and humidities and the heat, mass and momentum transfer
rates from the droplets to the gas phase are calculated, and the components
of the gas velocities are recalculated with these rates until convergence occurs.
This is known as the discrete droplet model or ‘‘particle-source-in-cell’’
concept. It can be seen that the principle is similar to that of the incremental
model, but in a much finer grid in three dimensions and with backward
influences from each cell to all those around, unlike the one-way stepwise
calculation of the incremental model.
Because the small errors in any calculation could build cumulatively,
it is essential to validate CFD and other complex theoretical models against
real physical measurements. SPS and co-workers, including Roger Keey,
showed that CFD could be successfully applied to spray dryers both
through its club funded research programme and, with industrial collabora-
tors (including ICI and Unilever), through a CEC-funded BRITE project
(Livesley et al., 1992). In these projects, detailed measurements of air flow
and droplet motion in a variety of spray dryers were obtained using phase-
Doppler anemometry (PDA) and used to validate CFD predictions. Goldberg
(1987) and Oakley et al. (1988, 1991) simulated a small cocurrent dryer using
the program [CFX-]FLOW3D (the forerunner of the modern CFX-4).
Experimental measurements of the air flow showed that increasing the
swirl vane angle from 25 to 30 to the vertical gave major changes in flow
pattern, and the simulation predicted this. Livesley et al. (1992) performed
similar validations on a number of industrial spray dryers. Oakley and Bahu
(1991) predicted a precessing vortex core at high swirl, and this effect was
confirmed by experimental observation (Langrish et al., 1993). Considerable
further work on this has taken placed recently, and the use of the more
modern program CFX5, which handles time-dependent flows better, has
given much more detailed results on these effects (LeBarbier et al., 2001).
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1736 KEMP AND OAKLEY


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 16. Typical particle motion problems encountered in spray dryers.

A detailed and up-to-date review of the theory and successful applications of


CFD in spray drying is given by Langrish and Fletcher (2001). CFD is now
proven as an invaluable tool in trouble-shooting, research and evaluation of
novel spray dryer designs.
Figure 16 shows a number of common spray dryer problems caused by
the particles taking an undesired path. All of these can be investigated using
CFD and low-cost solutions can then be assessed within the normal operat-
ing window. For example, a classic problem in the operation of spray dryers
is wall and roof impingement. Investigations using CFD reveal whether this
can be eliminated by altering the spray chamber aerodynamics by, for
example, moving the air entry point or adjusting air inlet velocities.
Wall and roof deposits are often caused by air pumping in rotary
atomisers, in which the high speed rotation of the atomizer wheel draws
in a stream of air and then expels it at high velocity with the atomized feed
stream. Air pumping can create an upward-flowing recirculation zone and
can cause a number of other problems such as low product density. Oakley
and Keey (1996) used CFD allied with experimental techniques to quantify
the effect.
In other case studies, problems with particles being swept rapidly out of
the dryer in the fast-flowing core have been overcome by increasing inlet air
swirl, and overheating of particles (due to dry particles recirculating into the
hot air inlet stream) was cured by altering the air entry point and changing the
inlet flow patterns. Previously, the only possible approach to these problems
associated with spray chamber aerodynamics was to use large-scale
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1737

trial-and-error, by altering gas flows and temperatures in the hope of finding


an improvement. Masters (1996) describes how leading spray dryer manufac-
turers are now using CFD techniques to avoid wall deposits and other product
quality problems in spray dryers. They can also explore novel types of spray
dryer, as they can now predict performance without having to build a full-size
test dryer.
Drying kinetics measurement still presents a major unsolved challenge.
It would be very useful to know the point at which the crust is likely to form
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

on the droplet, as this affects not only the kinetics but the morphology of the
resulting particle. Droplets have been suspended on wires or optical fibres
(Audu, 1978; Cheong et al., 1986; Huber and Keey, 1984) and the deflection
has been measured, but it has proved difficult to get the droplet small
enough—minimum sizes are 1 mm, as against the 30–200 micron droplets
produced from typical atomisers. Conceptually, a possible solution is to fire
droplets down a long, narrow tube and collect the product at the exit, but
even then it will be difficult to monitor the droplets at intermediate points.

APPROPRIATE MODELS FOR LAYER DRYERS

The most common method of handling these is to use the scaling


(integral) model for drying kinetics, using the characteristic drying curve
and the Specific Drying Rate concept (Moyers, 1994). The incremental
model can also be used to determine local conditions at different levels in
a layer or bed. Existing methods based on measuring batch drying curves
and scaling up have been reasonably successful, so less work has been done
in this area than on convective dryers and in particular dispersion dryers.
The drying kinetics can be measured on a suitable small-scale equivalent
of the dryer. For example, an experiment using a single tray in an oven can be
used to simulate tray dryers, continuous turbo-tray dryers, band dryers or
agitated dryers. The experimental conditions must be carefully chosen to
match the full-scale dryer in essential details. For example, the mode of heat-
ing must be the same (convective cross-circulation or through-circulation,
conductive at atmospheric pressure or under vacuum), the temperatures
used should be equivalent, and the same bed depth should be used. There
are also dryer-specific requirements. For example, a single tray being used to
determine the kinetics of a cross-circulation tray dryer should have the same
tray length as the full-scale dryer, so that the boundary layer flow and
Reynolds numbers are equivalent. One of the biggest problems is trying to
make a realistic simulation of agitation. The periodic turnover in plate, turbo-
tray and multiple band dryers can be simulated by periodically stirring the
material in a tray. For the continuous mixing in rotary and mechanically
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1738 KEMP AND OAKLEY

agitated dryers, a small-scale equivalent of the dryer is desirable (a stirred pot,


a Rototherm laboratory evaporator-dryer, or the MiMiPro lab-scale
microwave filter-dryer test unit developed by Pro-C-epT n.v.). Ultra small-
scale tests include the use of TGA (thermogravimetric analysis), allowing a
measurement on a few grams or even a single particle, a major advantage for
high-value pharmaceuticals.
When the drying kinetics have been measured, scale-up is a compara-
tively simple process. The SDR as defined by Moyers (1994) is the mass of
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

dry solids divided by the heated surface area and the drying time, and thus is
a mass flux.
m
SDR ¼ S ð30Þ
AS 
Moyers (1994) gives a number of alternative formulations of the SDR
in terms of mass flux, mass of bed, drying time etc., and states that if the
conditions of the small- and large-scale experiments are truly equivalent,
the SDR should be the same for both the test and the full-scale dryer.
This implicitly assumes that the characteristic drying curve concept applies.
In essence, the SDR concept is similar to the scoping design method, but it
makes allowance for falling-rate drying kinetics by using actual experimen-
tally measured kinetics. The size of the full-scale dryer can then be determined
from the relationship:
Afull mfull test
¼ ð31Þ
Atest mtest full
So for the full-size dryer, if the drying time is to be the same as for the
test, the surface area will have to be increased in proportion to the change in
mass. For many batch dryers, this is physically impossible, as the surface-
area-to-volume ratio falls as diameter is increased. Hence the drying time
increases on scale-up, as is known to occur in practice.
The justification for the SDR concept can be seen from a simple heat
balance or heat transfer equation, assuming no heat losses and no net heating
of solids:
 
dX
QP ¼ KAS T ¼ mS ev ð32Þ
dt
Summing over a given drying period to get a time  and a change of
moisture content X, this rearranges to:
mS KT
¼ ð33Þ
AS  Xev
The left-hand term is the SDR; the right-hand side includes a driving
force, the change in moisture content and the latent heat of evaporation.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1739

If the experiments were equivalent, and the initial and final moisture content
are the same for the test and full-scale dryer, then the SDR must be the same.
If the driving forces are different, we can formulate scaling rules. There are
two situations where this applies:
(a) Where the drying is heat balance controlled and the drying rate is
proportional to the total heat input. This occurs in a through-
circulation convective dryer where the airflow is proportional to
the cross-sectional area. The T term will be the temperature
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

drop between inlet and outlet, and the coefficient K will be a


heat capacity. This also applies for fluidised beds (and explains
the mB/A term in the equations).
(b) Where the drying is heat transfer controlled and the drying rate is
proportional to the heat transfer rate. This will be the case in a
contact dryer and some convective dryers. The T term will be
the temperature difference between heat source and solids, and
the coefficient K will be a heat transfer coefficient.
Moyers (1994) showed how changes in operating conditions during the
cycle could be handled by measuring a number of drying curves at different
conditions and breaking the design curve into sections. However, this may
be inconvenient or impossible. If the full-scale conditions cannot be
replicated in the small-scale test, but the change in external driving force
(heat transfer or rate of heat supply) is known, the analysis above can be
used to scale the evaporation rates to the new operating conditions, in a
similar way to that for fluidised beds.
For contact dryers a suitable scaling parameter should be (TW  TS),
where TW is the wall temperature (or temperature of the heating medium) and
TS is the solids temperature (or, for vacuum drying, the boiling point of the
solvent at the applied pressure). This is the driving force for heat transfer. For
convective dryers with low NTUs, the scaling factor should be (TGI  Twb),
where TGI is the inlet gas temperature and Twb is the wet bulb temperature of
the solids. Again, this is a heat transfer driving force. For convective dryers
with high NTUs, the scaling factor should be (TGI  Tas), where Tas is the
adiabatic saturation temperature. This is based on a heat and mass balance,
with all the heat available from the gas being lost as it passes through the layer.
All these scaling factors are based on external heat transfer or heat
supply, but most drying curves are dominated by the falling-rate period.
Hence, they will only apply if the characteristic drying curve concept
is obeyed, so that hindered (falling-rate) drying scales in proportion to
unhindered drying.
Detailed models are available for predicting the major parameters,
e.g., heat transfer coefficients h in contact dryers (Schlünder, 1984;
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1740 KEMP AND OAKLEY

Schlünder and Mollekopf, 1984). Only limited evidence exists that these
methods can successfully predict h from first principles for non-spherical
particles, wide size distributions and sticky materials. However, they should
be able to predict trends as temperature, agitation rate and other external
parameters change, based on an initial experimental value of h. Although
falling-rate drying kinetics is not considered, it will again be possible to
estimate new drying times by scaling the external heat transfer rate,
making the implicit assumption that the CDC concept applies.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

BEYOND UNIT OPERATIONS—THE


OVERALL PROCESS

It is important to consider the dryer as part of the overall system,


including upstream and downstream processing. The traditional approach
to process analysis is by unit operations: separating crystallization, filtration,
drying, solvent extraction, etc. Although it can be valuable to break the
problem down into manageable entities, this has the drawback of not
showing how one process affects another.
The crystallization or solids formation step has the main effect on
particle morphology, including particle size distribution, shape, and attrition
resistance. These parameters affect the downstream processing; in particular,
finer particles are harder to wash and dewater, and often make downstream
handling much more difficult. The effects are vital in the selection of dryers
(Kemp, 1999). The following diagram, Fig. 17, generated by Price (1998),
illustrates how the different steps in a process (e.g., a pharmaceutical
manufacturing process), from chemical synthesis to final micronisation and
tableting, affect the major parameters of importance for process behaviour
and final product specification. The relatively new field of product engineer-
ing gives opportunities to explore the interactions between the different unit
operations, to consider their effect on final particle properties, and to see how
this affects selection, design and problem-solving in dryers and in an overall
solids process.
A ‘‘layering’’ approach is helpful when considering how to model the
overall process. The principal selection choices and interactions between
different process steps need to be determined at a relatively early stage; limited
data is available and weak interactions will not affect the global optimum
significantly. Hence, heat and mass balances and scooping design calculations
(to supplement dryer selection by giving a rough sizing, an estimate of air
flowrate and the approximate dryer ‘‘footprint’’) are useful, but scaling and
detailed design models are not required at this stage.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1741


Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Figure 17. Stages of a typical pharmaceutical process and the parameters affected
at each step. Key: NCE ¼ new chemical entity, CSD ¼ crystal size distribution,
SLS ¼ solid–liquid separation.

In troubleshooting, simple heat and mass balances often give major


clues for the reason for the problem. As described by Kemp and Gardiner
(2001), basic models can reveal whether the root cause is inadequate heat
supply, drying kinetics (including insufficient residence time) or equilibrium
moisture content. Conversely, the problem may have its roots in a highly
localised region of the dryer, so that a detailed model is required to identify
the precise conditions at that point.

CONCLUSION

For decades, the Holy Grail of drying theory has been a ‘‘grand
unified theory of drying’’, as described by Piet Kerkhof in his final plenary
paper in the 2000 International Drying Symposium at Noordwijkerhout,
and discussed with some vigour in the ensuing questions! (see the summary
by Keey, 2001). Since Luikov and Whitaker’s analyses, people have
searched for the grand unified theory by detailed modelling of moisture
transport processes within solids. However, it has proved elusive because
of the complexity of the theory and the difficulty in measuring accurately the
many different parameters for each material that would allow the equations
to be solved.
Yet perhaps we do already have a unified theory of drying, except that
it is at a lower theoretical level and makes significant use of experimental
measurements to provide the necessary information. It is based on lumped
parameters, not distributed ones; scale-up from pilot-plant and lab-scale
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1742 KEMP AND OAKLEY

experiments, rather than design and simulation from basic physical


properties data alone. But it has proved successful in design, debottleneck-
ing and troubleshooting of all different types of dryers the world over, and
has been successfully validated against experimental results both on the
small and large scale. Where simplified theory is used, it is based as closely
as possible on the rigorous underlying theory, not empirical correlations
which are not capable of extrapolation because their format is not based
on real physical phenomena. Rigorous and complex methods such as CFD
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

are used where appropriate, but unnecessary complexity is avoided. This is


the legacy that Roger Keey and the many other researchers in drying have
left to us as we move into the 21st century.

NOTATION

A Surface area (m2)


a Dense phase velocity number (–)
a Surface area per unit volume or mass? (m2)
CP Specific heat capacity (J/kgK)
D Diameter of dryer (m)
d Differential operator (dt, dz) (–)
dP Particle mean diameter (m)
e Exponential constant, 2.71828. . . (–)
f Falling-rate drying factor (–)
G Mass velocity of gas (mass flux, or mass flow per unit area)
(kg/m2s)
g Acceleration due to gravity, approximately 9.81 m/s2 (m/s2)
H Enthalpy (J/kg)
hPG Heat transfer coefficient, gas to particles (W/m2K)
J Mass flux (of evaporated vapour) (kg/m2s)
K Coefficient in Eq. (32); heat transfer or specific heat
[context-dependent]
Kfl Airborne/dense phase time ratio parameter (–)
kY Mass transfer coefficient (based on humidity driving force)
(kg/m2s)
L Length of dryer (m)
m Mass (of bed or particle) (kg)
N Rotational speed of dryer (1/s)
N Drying rate (rate of change of moisture content) (kg/kg/s)
Ncr Unhindered drying rate at critical moisture content
(kg/kg/s)
NTU Number of transfer units (–)
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1743

Nu Nusselt number, hPG dP/ (–)


Pr Prandtl number, CP / (–)
Q Heat flow rate or rate of heat transfer (W)
Re Reynolds number, G U dP/ (–)
SDR Specific drying rate (kg/m2s)
T Temperature (C)
t Time (s)
U Velocity (m/s)
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

W Mass flowrate (dry basis) (kg/s)


X Solids moisture content (dry basis) (kg/kg)
Y Absolute humidity (mixing ratio) of gas (dry basis) (kg/kg)
z Axial distance (m)
Z Normalisation factor (–)

Greek Letters

Slope of rotary drum axis from horizontal (radians)


 Difference operator (–)

Angle from horizontal where solids fall from flight
(radians)
 Thermal conductivity (W/mK)
l Latent heat of evaporation (J/kg)
Viscosity (kg/ms)
 Factor allowing for humidity not being a true driving
force (–)
Circular constant, 3.14159. . . (–)
 Density (kg/m3)
 Solids residence time or required drying time (s)

Subscripts and Superscripts

0 At datum conditions
1 In airborne phase
2 In dense phase
as At adiabatic saturation conditions
B For bed of material
ev For complete evaporation
full For full scale dryer
G For gas
I At inlet
L For liquid/moisture
lm Log mean (temperature difference)
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1744 KEMP AND OAKLEY

O At outlet
P For particles
S For solids
test For small scale test
wb At wet bulb conditions
W At wall or heated surface
Wl Wall heat losses
Y For vapour
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

ACKNOWLEDGMENT

A large amount of the research and development of theory and


software described in this paper, both by AEA Technology and by external
researchers, was funded as part of the work programme of SPS (Separation
Processes Service, Harwell, UK) whom the authors thank for permission to
publish this paper.

REFERENCES

1. AFSIA. Cahier No. 19, Sechage et Aeraulique, Proceedings of Meeting


on March 22–23, 2001, Poitiers, France, AFSIA-ESCPE, 69616;
Villeurbanne, cedex, 2001.
2. Audu, T.O.K. Spray Drying Characteristics of Various Detergent
Slurry Drops. Tenside Detergents 1978, 15(1), 13–15.
3. Bahu, R.E. Fluidised Bed Dryer Scale-up. Drying Technol. 1994,
12(1–2), 329–339.
4. Baker, C.G.J. Cascading Rotary Dryers, Chapter 1. In Advances in
Drying; Mujumdar, A.S., Ed.; Hemisphere: New York, 1983; Vol. 2,
1–51.
5. Baker, C.G.J. The Design of Flights in Cascading Rotary Dryers.
Drying Technol. 1988, 6(4), 631.
6. Baker, C.G.J. Air–Solids Drag in Cascading Rotary Dryers. Drying
Technol. 1992, 10(2), 365–393.
7. Burgschweiger, J.; Groenewold, H.; Hirschmann, C.; Tsotsas, E.
From Hygroscopic Single Particle to Batch Fluidized Bed Drying
Kinetics. The Canadian Journal of Chemical Engineering 1999, 77,
333–341.
8. Cao, W.F.; Langrish, T.A.G. Comparison of Residence Time Models
for Cascading Rotary Dryers. Drying Technol. 1999, 17(4&5), 825.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1745

9. Cao, W.F.; Langrish, T.A.G. The Development and Validation of a


System Model for a Countercurrent Cascading Rotary Dryer. Drying
Technology 2000, 18(1&2), 99–116.
10. Cheong, H.W.; Jeffryes, G.V.; Mumford, C.J. A Receding Interface
Model for the Drying of Slurry Droplets. AIChEJ 1986, 32, 1334–1346.
11. Crowe, C.T. Modelling Spray–Air Contact in Spray-Drying Systems.
In Advances in Drying; Mujumdar, A.S., Ed.; Hemisphere: New York,
1980; Vol. 1, 63–99.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

12. Friedman, S.J.; Marshall, W.R. Studies in Rotary Drying. Part 1,


Holdup and Dusting. Chem. Eng. Progr. 1949, 45(8), 482–493.
13. Froessling, N. Gerlands Beitr. Geophys. 1938, 52, 170; Quoted in
Keey, 1972, p. 163.
14. Fyhr, C.; Kemp, I.C. Mathematical Modelling of Batch and
Continuous Well-Mixed Fluidised Bed Dryers. Chemical Engineering
and Processing 1999, 38, 11–18.
15. Fyhr, C.; Kemp, I.C.; Wimmerstedt, R. Mathematical Modelling of
Fluidised Bed Dryers with Horizontal Dispersion. Chemical Engineer-
ing and Processing 1999, 38, 89–94.
16. Fyhr, C.; Kemp, I.C. Evaluation of the Thin-Layer Method Used for
Measuring Single Particle Drying Kinetics. Transactions IChemE,
Part A 1998a, 76: A7 (October 1998) 815–822.
17. Fyhr, C.; Kemp, I.C. Comparison of Different Kinetics Models for
Single Particle Drying Kinetics. Drying Technology 1998b, 16(7),
1339–1369.
18. Gardiner, S.P. Dryer Troubleshooting. SPS Drying Manual Volume XI
(Dryer Operations) Part 3; Available only to licencees of the SPS
Process Manual, 1996.
19. Goldberg, J.E. Prediction of Spray Dryer Performance, D.Phil. Thesis;
University of Oxford, 1987.
20. Hirschmann, C.; Fyhr, C.; Tsotsas, E.; Kemp, I.C. Comparison of Two
Basic Methods for Measuring Drying Curves: Thin Layer Method
and Drying Channel, Proc. 11th Int. Drying Symp., Halkidiki, Greece,
1998; Vol. A, 224–231.
21. Huber, R.A.; Keey, R.B. Spray Dryers. SPS Drying Manual Volume V
Part 3; Available only to licencees of the Process Manual or subscribers
to SPS, 1984.
22. Keech, A.M.; Keey, R.B.; Kemp, I.C. A New Apparatus to Determine
the Drying Behaviour of Particulate Solids at Low Moisture Contents.
Drying 96, Proc. 10th Int. Drying Symp., IDS’96, Krakow, Poland,
July 30–Aug 2, 1996, Polish Acad. of Sciences et al. (Spons.);
Strumillo, C., Pakowski, Z., Eds.; ISBN 83 86903-07-04, 1996;
Vol. A, pp. 145–150.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1746 KEMP AND OAKLEY

23. Keech, A.M.; Keey, R.B.; Kemp, I.C. The Fundamental Mechanisms of
Drying Particles to Low-Moisture Contents; World Congress on
Particle Technology: Brighton, July 1998.
24. Keey, R. Drying; Principles and Practice; Pergamon Press, 1972.
25. Keey, R.B. Drying of Loose and Particulate Materials; Hemisphere:
New York, 1992.
26. Keey, R. Conference Report on the 12th International Drying
Symposium, IDS 2000. Drying Technology 2001, 19(1), 237–244.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

27. Kelly, J.J. Analysis of Design and Operating Procedures for Rotary
Dryers and Coolers. Part II. Introductory Analysis of Residence Time
Relationships. Technol. Ireland 1969a, 1(1), 15.
28. Kelly, J.J. Analysis of Design and Operating Procedures for Rotary
Dryers and Coolers. Part III. Effect of Gas Flow on Residence Times.
Technol. Ireland 1969b, 1(2), 25.
29. Kelly, J.J.; O’Donnell, P. Residence Time Model for Rotary Drums.
Trans. IChemE 1977, 55(4), 243.
30. Kemp, I.C. Progress in Dryer Selection Techniques. Drying
Technology 1999, 17(7&8), 1667–1680.
31. Kemp, I.C.; Bahu, R.E.; Oakley, D.E. Pneumatic Conveying Drying,
Proceedings of Pneumatech 4, Glasgow, 1990; (June 1990), pp. 413–426.
32. Kemp, I.C.; Bahu, R.E.; Oakley, D.E. Modelling Vertical Pneumatic
Conveying Dryers. Drying ’91, 7th Int. Drying Symp. Prague,
Czechoslovakia, Aug 1990; Mujumdar, A.S., et al., Eds.; Elsevier,
ISBN 0444893520, 1991; 217–227.
33. Kemp, I.C.; Bahu, R.E.; Pasley, H.S. Model Development and
Experimental Studies of Vertical Pneumatic Conveying Dryers.
Drying Technology 1994, 12(6), 1323–1340.
34. Kemp, I.C.; Fernandez, M. Unpublished SPS research 2001.
35. Kemp, I.C.; Frankum, D.P.; Abrahamson, J.; Saruchera, T. Solids
Residence Time and Drying in Cyclones. Drying’98, Proceedings
11th IDS, Thessaloniki, 1998; Vol. A, 581–588.
36. Kemp, I.C.; Fyhr, C.; Laurent, S.; Roques, M.; Groenewold, C.;
Tsotsas, E.; Sereno, A.; Bonazzi, C.; Bimbenet, J.; Kind, M. Methods
for Processing Experimental Drying Kinetics Data. Drying Technology
2001, 19(1), 15–34.
37. Kemp, I.C.; Gardiner, S.P. An Outline Method for Troubleshooting
and Problem-Solving in Dryers. Drying Technology 2001, 19(8),
pp. 1875–1890 (also in Proceedings 12th IDS, 2000, Noordwijkerhout,
[on CD-ROM], paper 271.).
38. Kemp, I.C.; Milborne, R. Design of Rotary Driers and Their
Application in the Fertiliser Industry; Proceedings of the International
Fertiliser Society, Paper no. 451, 2000.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1747

39. Kemp, I.C.; Oakley, D.E. Simulation and Scale-up of Pneumatic Con-
veying and Cascading Rotary Dryers. Drying Technology 1997, 15(6–8),
1699–1710 (also in Proceedings of 10th IDS, 1996, A, pp. 250–258).
40. Kemp, I.C.; Oakley, D.E.; Bahu, R.E. Computational Fluid Dynamics
Modelling of Vertical Pneumatic Conveying Dryers. Powder Technol.
1991, 65(1–3), 477–484.
41. Kunii, D.; Levenspiel, O. Fluidization Engineering, 2nd Ed.;
Butterworth-Heinemann: Boston, USA, 1991.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

42. Langrish, T.A.G. The Mathematical Modelling of Cascading Rotary


Dryers, D.Phil. Thesis; Oxford, 1988.
43. Langrish, T.A.G. Wind-Tunnel and Pilot-Plant Tests for Estimating
Entrainment in Cascading Rotary Dryers. SPS Research Note
Presented at SPS Drying Panel, January 1992.
44. Langrish, T.A.G. Flowsheet Simulations and the Use of CFD
Simulations. In Drying Technology, Drying ’96, Proc. 10th. Int.
Drying Symp. (IDS’96), Krakow, Poland, Lodz Technical University
(publ.); Strumillo, C., Pakowski, Z., Eds.; 1996; Vol. A, pp. 40–51.
45. Langrish, T.A.G.; Bahu, R.E.; Reay, D. Drying Kinetics of Par-
ticles from Thin Layer Drying Experiments. Drying ’91, (7th
Int. Drying Symp., Prague, Czechoslovakia, Aug 1990); Mujumdar,
A.S., et al., Eds.; Elsevier, ISBN 0444893520, 1991a; 196–206.
46. Langrish, T.A.G.; Bahu, R.E.; Reay, D. Drying Kinetics of Particles
from Thin Layer Drying Experiments. Chem. Eng. Res & Des.
(Transactions IChemE) 1991b, 69(5), 417–424.
47. Langrish, T.A.G.; Fletcher, D.F. Spray Drying of Food Ingredients
and Applications of CFD in Spray Drying. 2001. Paper submitted to
Trans. IchemE.
48. Langrish, T.A.G.; Oakley, D.E.; Keey, R.B.; Bahu, R.E.; Hutchinson,
C.A. Time-Dependent Flow Patterns in Spray Dryers. Trans. I. Chem.
E. 1993, 71(A), 355–360.
49. Langrish, T.A.G.; Papadakis, S.E.; Baker, C.G.J. Residence Times of
Two- and Three-component Mixtures in Cascading Rotary Dryers.
2001. Paper accepted by Drying Technology.
50. Langrish, T.A.G.; Reay, D.; Bahu, R.E.; Whalley, P.B. An Investiga-
tion into Heat Transfer in Cascading Rotary Dryers. J. Separ. Proc.
Technol. 1988, 9, 15.
51. LeBarbier, C.; Kockel, T.K.; Fletcher, D.F.; Langrish, T.A.G.
Experimental Measurement and Numerical Simulation of the Effect
of Swirl on Flow Stability in Spray Dryers. 2001. Paper accepted for
Chemical Engineering Progress.
52. Livesley, D.M.; Oakley, D.E.; Gillespie, R.F.; Elhaus, B.;
Ranpuria, C.K.; Taylor, T.; Wood, W.; Yeoman, M.L. Development
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1748 KEMP AND OAKLEY

and Validation of a Computational Model for Spray–Gas Mixing in


Spray Dryers. In Drying ’92, Proc. 8th Intl. Drying Symp. (IDS ’92),
Montreal, Canada; Mujumdar, A.S., Ed.; Elsevier, 1992; 407–416.
53. Luikov, A.V. The Drying of Peat. Ind. Eng. Chem. 1935, 27, 40–69.
54. Luikov, A.V. Heat and Mass Transfer in Capillary-Porous Bodies;
Pergamon Press: Oxford, UK, 1966.
55. McKenzie, K.A.; Bahu, R.E. Material Model for Fluidised Bed
Drying. In Drying ’91 (Selected papers from 7th Int. Drying Symp.,
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

Prague, Czechoslovakia, Aug 1990); Mujumdar, A.S., et al., Eds.;


Elsevier, ISBN 0444893520, 1991; 130–141.
56. Masters, K. Deposit-Free Spray Drying: Dream or Reality? In
Drying ’96, Proc. Tenth International Drying Symposium (IDS ’96),
Strumillo, C.; Mujumdar, A.S., Eds.; Krakow, Poland, 1996; Vol. A,
52–60.
57. Matchett, A.J.; Baker, C.G.J. Particle Residence Times in Cascading
Rotary Dryers. Part 1—Derivation of the Two-Stream Model. Journal
of Separation Process Technology 1987, 8, 11–17.
58. Matchett, A.J.; Baker, C.G.J. Particle Residence Times in Cascading
Rotary Dryers. Part 2—Application of the Two-Stream Model
to Experimental and Industrial Data. J. Separ. Proc. Technol.
1988, 9, 5.
59. Matchett, A.J.; Sheikh, M.S. An Improved Model of Particle Motion
in Cascading Rotary Dryers. Trans. IChemE 1990, 68(Part A), 139.
60. Moyers, C.G. Scale-up of Layer Dryers; A Unified Approach. Drying
Technology 1994, 12(1&2), 393–416.
61. Oakley, D.E. Design of Layer and Contact Dryers. SPS Drying
Manual Volume VII Part 4; Available only to licencees of the SPS/
Process Manual, 2001.
62. Oakley, D.E.; Bahu, R.E.; Reay, D. The Aerodynamics of Cocurrent
Spray Dryers. In Proc. 6th Int. Drying Symp, IDS ’88; Roques, M.,
Ed.; Versailles, France, OP, 1988; 373-OP 378.
63. Oakley, D.E.; Bahu, R.E. Spray/Gas Mixing Behaviour within Spray
Dryers. Drying ’91; Mujumdar, A.S., Filkova, I., Eds.; Elsevier:
Amsterdam, 1991; 303–313.
64. Oakley, D.E.; Keey, R.B. Air Pumping by Rotary Atomizers in Spray
Dryers. Drying 96, Proc. 10th. Int. Drying Symp. (IDS’96), Krakow,
Poland; Strumillo, C., Pakowski, Z., Eds.; ISBN 83 86903-07-04, 1996;
Vol. A, pp. 513–520.
65. Papadakis, S.E.; Langrish, T.A.G.; Kemp, I.C.; Bahu, R.E. A
Short-cut Design Method for Cascading Rotary Dryers. Drying ’92,
Proceedings of 8th International Drying Symposium, Montreal, 1992;
pp. 1258–1267.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

MODELLING OF PARTICULATE DRYING 1749

66. Price, C.J. Developing a Reliable Crystallization Process. SPS


Symposia, Maidenhead, UK and Wilmington, USA, 1998; Available
only to subscribers to SPS, 1998.
67. Ranz, W.E.; Marshall, W.R. Evaporation from Drops. Chem. Eng.
Progr. 1952, 48(3), 141–146 and 48(4), 173–180.
68. Reay, D. A Scientific Approach to the Design of Continuous Flow
Dryers for Particulate Solids. Multiphase Sci. & Technol.; Hewitt,
G.F., Delhaye, J.M., Zuber, N., Eds.; Hemisphere Pub. Corp.: ISBN
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

0891166491, 1989, Vol. 4, pp. 1–102.


69. Reay, D.; Allen, R.W.K. Predicting the Performance of a Continuous
Well-mixed Fluid Bed Dryer from Batch Tests, Proc. 3rd Int. Drying
Symp., Birmingham; Ashworth, J.C., Ed.; Drying Research Ltd.,
1982a; Vol. II, 130–140.
70. Reay, D.; Allen, R.W.K. The Effect of Bed Temperature on Fluid Bed
Batch Drying Curves. J. Sep. Process Technol. 1982b, 3(4), 11–13.
71. Renaud, M.; Thibauld, J.; Trusiak, A. Solids Transportation Model
of an Industrial Rotary Dryer. Drying Technology 2000, 18(4&5),
843–865.
72. Saeman, W.C.; Mitchell, T.R. An Analysis of Rotary Dryer and
Cooler Performance. Chem. Eng. Progr. 1954, 50(9), 467.
73. Saruchera, T. Measurement and Modelling of Particle Residence Time
in a Return-Flow Cyclone, Ph.D. Thesis; University of Canterbury:
New Zealand, 1999.
74. Schofield, F.R.; Glikin, P.G. Rotary Dryers and Coolers for Granular
Fertilizers. Trans. IChemE 1962, 40, 183.
75. Schlünder, E. Heat Transfer to Packed and Stirred Beds
from the Surface of Immersed Bodies. Chem. Eng. Process. 1984, 18,
31–53.
76. Schlünder, E.; Mollekopf, N. Vacuum Contact Drying of Free-flowing
Particulate Materials. Chem. Eng. Process. 1984, 18, 93–111.
77. Shahhosseini, S.; Cameron, I.T.; Wang, F.Y. A Simple Dynamic
Model for Solid Transport in Rotary Dryers. Drying Technology
2000, 18(4&5), 867–886.
78. Tsotsas, E. From Single Particle to Fluid Bed Drying Kinetics. Drying
Technology 1994, 12(6), 1401–1426.
79. van Meel, D.A. Adiabatic Convection Batch Drying with Recircula-
tion of Air. Chem. Eng. Sci. 1958, 9, 36–44.
80. Vanecek, V.; Picka, J.; Najmr, S. Some Basic Information on the
Drying of Granulated NPK Fertilisers. Int. Chem. Eng. 1964, 4(1),
93–99.
81. Vanecek, V.; Picka, J.; Najmr, S. Fluidised Bed Drying; Leonard Hill:
London, 1966.
MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

1750 KEMP AND OAKLEY

82. Whitaker, S. Simultaneous Heat, Mass and Momentum Transfer in


Porous Media. Advances in Heat Transfer; Academic Press: New
York, USA, 1977; Vol. 13, 119–203.
83. Whitaker, S. Heat and Mass Transfer in Porous Media. Advances in
Drying; Mujumdar, A.S., Ed.; Hemisphere: Washington D.C., USA,
1980; Vol. 1, pp. 23–61.
84. Williams-Gardner, A. Industrial Drying; Leonard Hill Books: London,
1971.
Downloaded by [Moskow State Univ Bibliote] at 04:32 10 October 2013

You might also like