Professional Documents
Culture Documents
1994 - Kemp - Scale-Up of Pneumatic Conveying Dryers
1994 - Kemp - Scale-Up of Pneumatic Conveying Dryers
To cite this article: I C Kemp (1994) SCALE-UP OF PNEUMATIC CONVEYING DRYERS, Drying Technology: An International
Journal, 12:1-2, 279-297, DOI: 10.1080/07373939408959957
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
DRYING TECHNOLOGY, 12(1&2), 279-297 (1994)
1 C Kzmp
Downloaded by [McGill University Library] at 19:02 20 November 2014
ABSTRACT
Scale-up of pneumatic conveying dryers has until now been !:ugely ernpiric:ll.
A theoretical model is outlined which predicts dryer performance cffectively. It
depends on some uncertain parameters, norably agglomeration and wall friction, but
these can be found more accurately by checking the predictions against resulrs from
the small-scale dryer. The mode! can then be used to scale up to the lull size dryer.
The required duct length falls as the gas velocity is reduced md rhr duct diamr~er
increased, subject to satisfactory conveying and dispersion ar the t'erdpoin~.
INTRODUCTION
The concept of pneumatic conveying dryers, somedrnes known as flash dryers.
is simple; hot gas conveys a pmiculate solid along a duct and dries it as it transports
it. Figure 1 illustrates a typical unit. However, producing a theoretical mode! of even
reasonable accuracy has proved very difficult in the past. As a result, manufacturers
have universally designed pneumatic conveying dryers by using pilot-plant tests to
evaluate the drying rate of the material and the conditions required, and then scaling
up to the full-size dryer. The scale-up procedures used have been empirical in nature,
and have varied considerably between different manufacturers. Clearly, a reliable
scale-up procedure with a clear theoretical basis is highly desirable. This paper aims
to suggest such a procedure.
The theoretical design procedures have been much improved in recent years.
However, the dryer behaviour is complex and it is very difficult to measure certain
parameters which affect drying rates considerably, such as agglomeration. Hence, it
EXHAUST
AIR
B A G FILTER
Steady -state
Downloaded by [McGill University Library] at 19:02 20 November 2014
,
zone
I
DRY SOLIDS
OUTLET
WET SOLIDS
INLET
::.. Acceleration
zone
- INLET
AIR
is unlikely that pneumatic conveying dryers can be designed from theory alone in [he
foreseeable future; pilot plant tests willcontinue to be necessary. However, the theory
allows us to understand the rules that will apply to scale-up, and the uncertain
parameters can be checked by comparing theoretical predicdons with the pilot-plant
results - a "fitting mode calculation".
DRYER DESIGN MODELS
Historical models
For the basic case of a vertical drying duct, particles enter at h e feedpoint, are
entrained in the fast-moving gas saeam and are accelerated till they reach a steady-
state velocity. The gas-to-particle relative velocity is then similar to the particle
termind velocity. Heat is transferred from the hot gas to the pardcles by convection.
This heat is used to evaporate liquid from the particle surface and any surplus heat
warms up the particle. Hence a full analysis must consider panicle motion, heat
transfer and mass transfer.
PNEUMATIC CONVEYING DRYERS 28 1
that of Thorpe. Wint and Coggan2.However, the estimated duct lengths for a given
drylng duty could still be 50-100% in error. Consequently these models have nor
replaced the mditional procedure for dryer design of running pilot-plant tests and
then scaling up using empirical parameters.
The scale-up methods of individual manufacturers have not been published, but
suggested values for the increase in duct length required for an order of magnitude
throughput increase have ranged from zero to 50-100%. Of the very few theoretical
scale-up factors proposed in the open literature, none takes any account of the
acceleration zone and any agreement with actual results will be largely coinciden~al.
Formulation of the SPS model
The philosophy of dryer modelling and design has been discussed by Renyl.
Models for dryers can be divided into two pans:
The equipment model examines factors which are specific to the kind of drying
equipment used. These include:
Gas flow patterns
Solids transpon
Heat transfer to solids
The material model examines factors specific to the material being dried:
Dryingkinetics
Drylng equilibria
The two parts are combined to give an overall drying model.
This approach aims to dlow accurate prediction of dryer design and performance
with little or no large-scale testing. Three main tools are required:
a good theoretical model of the equipment to allow prediction af the
drying processes
small-scale tests to measure the parameters required for the theoretical model
* pilot-plant testing to verify the model and allow fine tuning.
KEMP
Downloaded by [McGill University Library] at 19:02 20 November 2014
The theoretical model not only helps design from scratch, but also allows more
accurate scale-up to be cxried out from pilot plant data where this has previously
been carried out by empirical methods.
The overall model may be an integral model or an incremental model. The
integral model is simpler; it is described by ~ a h and
u ~works well for many crossflow
dryers, such as belt, band, through-circulation and fluidised bed dryers. Drying
kinencs can be measured in small-scale tests at constant gas remperarure and scale-up
to different conditions can often be based on a multiplying factor or dimensionless
group.
However, in dryers where the gas ahd solids are in cocurrent or countercurrent
flow, the gas in contact with the solids is at different temperatures and hupidiries at
different points in the dryer, altering the driving forces for hear and mass aansfer.
This makes it virtually impossible to carry out small-scale tesrs which will exactly
reproduce the temperature and humidity history of the particle in the full-scale dryer.
Hence for such dryers, which include cascading rotary dryers, pneumatic conveying
and spray dryers, a more detailed model is needed.
Incremental models
" ~ e a described
~ ' how co-c-nt and counter-current dryers can be modelled in
a one-dimensional incremental manner as shown in Figure 2 for the case of a
co-current unit. The dryer is divided into alarge number of steps. Starting at the inlet
end, calculation proceeds by increments until the outlet moisture content is achieved.
If we consider a small element of length dz then a series of differential equations can
be written to describe the processes occumng in the dryer. In the following equacions.
W is the dry mass flowrate, T is temperature, X is the dry-basis moisture content, Y
is the absolute humidity and C, is the specific heat capaciry; subscripr C denotes gas,
S dry solids, L liquid (moisture) and Y vapour.
PNEUMATIC CONVEYING DRYERS 283
Drying
where
h =heat uansfer coefficient from gas to solids (from equipment model)
A, = specific surface.area of solids per unit mass.
The function is found from the material model, and may be a simple correlation
Downloaded by [McGill University Library] at 19:02 20 November 2014
of the drying coefficient type or a full mechanistic mode1 including drying equilibria.
Heat transfer
dQ, is the wail heat loss term, which is normally either taken as s percentage
of the heat carried by the gas stream or calculated from appropriate correlations for
heat losses by natural or forced convection.
Rearranging and neglecting second order terms:
2 84 KEMP
4. Repeat steps 1-3 until the product moisture content has been attained.
This model can be used as the basis foranalysing most continuous dryers. Where
the gas and material flows are countercumnr or cocurrent, [he gas and solids
conditions at the exit from one increment are used as the entry conditions for rhe nexr
one. In crossflow dryers, the gas e n q and exit conditions for each increment fie
found individually and their mean values can be found by summing over all the
increments along the dryer.
Application to pneumatic conveying dryers
The incremental model is particularly useful for pneumatic conveying dryers,
which can be modelled as one-dimensiond cocurrent units by tracking the path of
the particle as it moves along the drying tube. A high propomon of the drying rakes
place in the region near the feedpoint, where gas temperature is highest and
particle-to-gas relative velocity is also at a maximum, so that the heat transfer
coefficients are high. Hence a very short step length is needed in this initial region,
and the number of steps needed for accurate modelling is beyond the possibilities of
hand calculation; a computer program is needed.
PNEUMATIC CONVEYING DRYERS 2 85
Here p ¬es density, subscript P is used for individual particles, and subscripr
Wis wet basis (i.e. including associated moisture). U , is the relative velocity between
gas and particles, equal to Vc-Up. The drag coefficient CDcan be obtained from the
relationships reponed by Clift, Grace and webe?. The force balance in the form
above is known as the "suspension model" and includes a particle-to-wall friction
term which depends on tube diameter D, particle velocity U pand a friction factor fp.
SPS experiments have c o n f i i e d that this term is significant, particularly for small
diameter ducts (less than about 0.2 m); this has particular implications for scde-up.
At steady state, dUp/dr=Oand the steady-state relative velocity (U,),, is given by the
equation below; wail friction can make (UR)SSmore than 1008 greater than [he
terminal velocity of a single particle U,.
Various coklations have been suggested forf,, mostly of the form fp = K,IUp ,
where K,is a (dimensional) constant. Different workers have reponed widely different
values of Kr Cornpaison has not been helped by the use of two different definirions
off,, one being four times the other; the definition used here is that advocated by
yangb,in terms of wall shear stress 7,:.
elasticity and wall roughness. Hence, only rough estimates ofKfandfpcan be obtained
from general principles; ideally, they should be measured experimentally. For most
systems, however, K,appears to be of the order of 0.2 m-'s.
Heat transfer
The most commonly used correlation for heat transfer to particles is due to Rsnz
and ~arshall',although it was developed for experiments on evaporation from water
droplets.
Downloaded by [McGill University Library] at 19:02 20 November 2014
Here ReR, is the Reynolds number appropriate for heat uansfer, which is based
on gas-to-particle relative velocity U, and on Sauter mean panicle diameter d,,,.
For large particles or high fluid velocities, a turbulent outer boundary layer also
develops, particularly in the wake of the panicle. webe? added an additional term
proportional to R ~ ~ O ' to the Ranz and Marshall correlation to account for turbulent
flow at particle Reynolds numbers greater than 130:
4. Remn the new model at the conditions for [he full-size plant and predict its
performance.
5. Check areas where the design is especially sensitive, such as the feedpoint. using
CFD or similar tools.
Let us consider each step in more depth.
Step 1 requires a pilot plant or exisang dryer with sufficient insnumenution.
Downloaded by [McGill University Library] at 19:02 20 November 2014
the panicle then simply heats up to the gas temperature. The solution is to change the
numbers used in the built-in drying kinetics model; for example, by increasing rhe
value given for the critical moisture content X,, which marks the transition between
unhindereddrying ("constant-rate") and hindered drying ("falling-rate") regimes. The
pro- is re-run until a good fit is obtained. It may also'be possible to measure the
drying kinetics more accmtely using a thin-layer test as described by Langrish, Bahu
and Reayl'. The very rapid drying makes this difficult with present techniclues, bur
development is continuing.
Downloaded by [McGill University Library] at 19:02 20 November 2014
If the observed drying is more than the prediction, this is almost cenainly due
to agglomeration. Recent SPS research has revealed the surprising fact that, for 3.
given gas veIocity, agglomeration of the solid feed actually tends to increase the
drying in a given length of duct. This is because the upward velocity and the heal and
mass aansfer rates both decrease for large agglomerates as compared with single
particles, but the increased residence time in any section of tube is found to ourweigh
the slower drying rates. Details of the model will be published in a forthcoming
paper1'.
An alternative possibility is that more drying than predicted is occumr~gin the
cyclone. Much less is known about panicle motion and heat and mass transfer in the
cyclone than in the main drying duct. Residence times can be compuablr or geatrr.
Although the gas and panicles are usually close to thermal equilibrium in the cyclone
and driving forces are low, drying can still continue, especially where there is :I
significant amount of moisture diffusing from within the panicles.
Step 4 requires data on the solids flownte and moisture content of the full-scale
system. The gas inlet temperature, gas velocity and duct diameter must also be
selected; a possible method is outlined in the next section. The simulation can then
be performed using thecomputerprogram. It is important toidenrify any orherchanges
between the pilot plant and the full-size one. For example, achange in the feed system
may change the degree of agglomeration or the distribution of moisture between the
particle surface and interior, affecting the results considerably.
Step 5 allows for some particular problems of large diameter ducts. Dispersing
the feed uniformly into the gas smearn is a particular problem with luge dryers and
a number of different feedpoint configurations have been tried, including venruris.
slingers, twin or multiple feedpoints and impinging sueams. Poor dispersion can
seriously reduce the drying occurring in the all-important acceleration zone.
Moreover, choking and recirculationeffects can be serious. It is advisable to use CFD
to check the design of the feedpoint region of a scaled-up dryer.
Implications for design
The model reveals many of the reasons why the empirical scale-up procedures
work. The empirical ruie that tube length should increase with diameter is shown to
be largely due to wall friction. Compared with a 20 rnrn or even 100 mm diameter
290 KEMP
pilot plant tube, panicles in a I m duct on the full-size plant will have higher panicle
velocity (reducing their residence time) and higher relative velocity (reducing the
heat transfer rate). Hence, to achieve a-given-amount of dry-tng, the ducr will need to
be longer than that on the pilot plant.
The significance of wall friction is revealed by Table I . This gives culculated
ducr lengths for drylng of 350 pm particles from a moisture content of 0.3 to 0.15
kg/kg (dry basis), with the solids-gas loading (phase density) held constanr at 0.1 kg
dry solids/kg dry gas. The table shows that the length of duct required to perform the
duty increases by 20-25% when scaling up from an experimental apparatus of 20 mm
tube diameter to a 1 mdiameter production plant, entirely due to wall friction effects.
It has been assumed that Kf=0.2. SPS has observed a few cases for which K,appears
to be considerably higher and the predicted increase in duct length on scale-up is
50-100%.
Overall, the wall friction term is proportional to Up.Hence, in the acceleration
zone, wall effects have only a minor effect on particle velocity. For very fast-drying
materials, e.g. small particles lightly wetted on the surface only, most of the drying
takes place in the acceleration zone. Little or no increase in tube length will then be
requiredduring scale-up. Drying kinetics experiments may be used to try to determine
the proportion of fast-drying surface moisture present. The use of the rheoreacal
model should allow more accurate scale-up than is provided by simply using an
empirical factor to increase tube length by 20-50%.
20 -..............-..-.-..-.................................................................................................
Blocking velocity for agglomerate
Downloaded by [McGill University Library] at 19:02 20 November 2014
I
I
L
t I I . I I I I
0
0 5 10 15 20 25 30 35
Gas velocity, m/s
-----
Single panicle Agglomerate
specification. this may again be cured by increasing the gas velocity. Table 1, however,
shows that when the gas velocity is increased, a longer tube is needed to achieve the
same amount of drying. Conversely, for a given length of tube, if the gas velocity is
increased, less drying is achieved. Further evidence is provided by Figure 3.
The plot in Figure 3 shows the tube length required to achieve a given drying
duty. The same 350 prn particles were being simulated; rube diameter was 0.857 m,
solids flowrate 0.5 kg/s (1800kgh),gas inlet temperature 150'C and moisture content
was being reduced from 0.3 kg/kg to 0.15 kglkg. It can be seen that, over virrually
the entire range, an increase in gas velocity necessitated an increase in duct length.
In other words, the decrease in pamcle residence rime due to the higher gas velocity
ourweighed the more rapid drying due to increased relative velocity between gas and
particles.
Calculaaons show that at a gas velocity U, of 2 nl/s (very close ro the terminal
velocity) the panicles cannot be conveyed up the tube and it blocks at the feedpoint.
It is not yet cleat whether this blocking velocity is the same as the choking velocity
reported and correlated by other workers; research on this is continuing. However,
KEMP
the simulation shows that uconsiderably higher velocity is required to make the dryer
work; as the gas velocity is decreased, the required tube length re~lches;i minimum
at UG = 4.5 d s and then climbs rapidly, reaching an asymptots nt 4.1 m/s. The
explanation is simple; the gas is running out of heat. As gas velocity decreases, the
gas mass flowrate decreases proponionally, as does its heat content. When this falls
below the constant heat load required to evaporate the moisture in the feed, the dryer
becomes infeasibte. So Uc should not be decreased too far. However, ir should be
noted that the optimum Ucis very close to the minimum feasible value U,,, at the
asymptote, and the latter is easily calculated by heat balance. The best dryer
Downloaded by [McGill University Library] at 19:02 20 November 2014
T,, can safely be increased, increasing the thermal efficiency of the dryer and
minimising the increase in gas flowrate and dryer size. However, if the additional
moisture is internally bound, *ere will be a danger of overheating if To is increased.
There is no merit in increasing the gas velocity above the minimum required
either to provide the heat needed toperform the required evaporation duty or to convey
agglomerates. In the latter case, high velocities are only inninsically necessary near
the feedpoint, as the agglomerates break up; hence a venturi is likely to be beneficial.
The one proviso is that the feed must disperse effectively across the larger diameter
Downloaded by [McGill University Library] at 19:02 20 November 2014
The outlet gas humidity Yo is found from an overall mass balance analogous to
equation (4):
8. If the dryer is very short and fat, its configuration can be changed by reruming
to step 5.
9. If the drying duct is impracticably long, consider the possibility of an cvrn
narrower venturi throat at the feedpoint to give a localised high velocity sufficient to
break up agglomerates. D for the main tube can then be increased and U, reduced,
which should give a shorter dryer. If this is not possible. a ring dryer will often be
the best solution.
It should be noted that this procedure is new and has not .been med by manu-
facturers in practice. However. tesdng on indusmd d u a is currently under way.
The problem of the cyclone /
The equations above only apply to the main drying duct. The scale-up of the
cyclone is still very much a black art.The cyclone is generally designed to keep solids
emissions down to an acceptable level, and little thought is given to the drying which
occurs there. In any case, there is no way of calculating it at present because there is
virtually no published information on solids residence time, heat uansfer and drying
rates.
SPS is currently beginning experimental work in this area, with the aim of
formulating a model similar to that for the main duct. One possibility is to treat the
cyclone as an exaa length of duct if the particle residence time can be estimated.
However, three clear differences from the main duct have been observed already.
Firstly, the particles do not move in plug flow; there appears to be considerable mixing
in the cyclone. Secondly, there is considerable solids-solids interaction, so that the
Ranz and Marshall heat amsfer correlation is unlikely to apply. Thirdly, the solids
are mainly in a layer very close to the wall; they will be in the boundary layer rather
than the main gas flow, and even solids-to-wall conduction could be significanr.
Experiments so far suggest that holdup of solids in the cyclone increases wirh
particle size and density and with gas velocity, presumably due to increased
centrifugal forces. One implication is that a reduction in gas velocity in the cyclone
may reduce the residence time and the drying achieved there, so that cyclone diameter
should be as small as possible - the opposite conclusion to that for the main duct.
However, many other factors come into play, including solids carryover, pressure
drop and particle attrition.
PNEUMATIC CONVEYING DRYERS
CONCLUSIONS
1. There is currently no consensus on scale-up methods for pneumatic conveying
dryers.
2. The SPS model offers a most promising method for scale-up with a fm
theoretical basis.
3. The model predicts the effect on drying rates of changing gas temperature and
velocity and other parameters, and thus allows the effect of design variations to
be calculated rapidly.
Downloaded by [McGill University Library] at 19:02 20 November 2014
4, The model predicts that minimising gas velocity will minimise dryer length,
subject to the constraints of the heat balance and of conveying the solids.
5. CFD is an effective tool for modelling flow patterns and particle dispersion near
the feedpoint.
6. Further practical testing of the model is highly desirable.
7. The major remaining uncertainties are; (a) the drying kinetics of small panicles
in high-velocity gas streams; (b) the cohesive forces and break-up mechanisms
in agglomerates; (c) the drying taking place in the cycIone. Research is required
in all these areas and is currently being performed by SLS:
-
NOMENCLATURE
A, Specific surface area of solids mZkg"
CD Drag coefficient of particle in gas s & e m
CpG Specific heat capacity of dry gas J kg'lK"
Cp, Specific heat capacity of liquid (moisture) J kg"~"
Cp, Specific heat capacity of dry solid I kg"K.'
Cp, Specific heat capacity of vapour J kg"K"
D Tube or duct diameter m
dp Particle diameter (on weight average basis) m
d,,,, Particle diameter (Sauter mean basis) m
fp Particle-to-wall friction factor (Yang definition)
fs Particle-to-wall friction factor (alternative definition) -
&! Acceleration due to gravity m s"
h Heat transfer coefficient between gas and particles W m'*K"
K, Factor in wall friction equation m-'s
Nu Nusselt number (based on particle diameter), Mp(SM,J~ -
Pr Prandtl number for gas, pCpJK
Qm Heat lost through walls of drying tube W
Re, Reynolds number for heat transfer basedon U, -
anddplsM1
TG Gas temperature (local value) 'C
Tc,TGo Gas temperature at inlet/outlet of drying tube 'C
Ts Particle temperature (local value) 'C
KEMP
dryers. In A.S. Mujumdar (ed), Drying '91. Paper originally presented at LDS
'90, Prague.
14. Oaktey, D.E. (1992). Scale-up of spray dryers with the aid of computational
fluid dynamics. Drying Technology, this issue.
15. Kemp, I.C., Oakley, D.E. and Bahu, R.E. (1991). Computational fluid dynamics
modelling of vertical pneumatic conveying dryers. Powder Technology, 65, pp.
477-484.