Modeling of Vacuum Contact Drying of Crystalline Powders Packed Beds

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Available online at www.sciencedirect.

com

Chemical Engineering and Processing 47 (2008) 722–730

Modeling of vacuum contact drying of


crystalline powders packed beds
A. Michaud, R. Peczalski ∗ , J. Andrieu
Université de Lyon, Université Lyon 1, ESCPE Lyon, CNRS, UMR 5007, Laboratoire d’Automatique et de
Génie des Procédés -LAGEP, Villeurbanne F-69616, France
Received 31 October 2006; received in revised form 18 December 2006; accepted 18 December 2006
Available online 23 December 2006

Abstract
Vacuum contact drying of monohydrate lactose and potassium chloride packed beds were experimentally investigated and numerically simulated.
The classical “vaporization front” model was improved to correctly represent the experimental drying rate data. The main model modification
consisted of introducing linear variation in solvent content at the vaporization front as opposed to classical modeling where this parameter was
assumed constant. The new model, as applied to the falling rate drying period, involved only one fitting parameter. The validation experiments
were conducted with a laboratory vacuum contact dryer.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Vacuum contact drying; Modeling; “Vaporization front”; Packed bed; Powder

1. Introduction To our knowledge, the most detailed experimental study on


packed bed vacuum contact drying was conducted by Moyne
Contact heating in a vacuum is widely used in the pharma- [5]. This author conducted some experiments with pure water
ceutical industry to dry granular products, which are sensitive to and 200 ␮m non-porous glass beads. According to Moyne, the
oxygen and temperature and are often toxic and explosive. The experimental drying rate curves presented four distinct periods:
major challenge in thermosensible powder drying is the preser-
vation of the initial grain size and shape during the process. • a short initial period during which both the bed temperature
The agglomeration and attrition phenomena, which depend pri- and drying rate increased;
marily on the stirring mode (rate, periodicity) and geometrical • a classical constant rate period during which the bed at the
parameters (stirrer, vessel), must be prevented. To succeed, the heating wall remained saturated with solvent;
easiest solution is to alternate periods of stirring and no stirring, • a first falling rate period, starting from a critical moisture
a process called contact drying under intermittent stirring. With practically independent of operating conditions;
this drying protocol, the study and analysis of stirred (agitated • a second falling rate period during which the slope of the
period) and packed bed (static period) drying phenomena are drying rate curve decreased to a greater extent.
equally important.
Whereas the contact drying of powder and grains in stirred Baillon [6] studied the experimental temperature profile dur-
bed has been investigated in detail by Schlünder and Mollekopf ing vacuum contact drying of pharmaceutical granular packed
[1], Tsotsas and Schlünder [2,3] and Farges et al. [4], little exper- beds. He observed that during the falling rate period, the temper-
imental data on packed bed drying kinetics in vacuum have been ature profiles inside the bed corresponded to two zones: a dried
published. zone near the heating wall, where the temperature gradient was
high and a wet zone where the temperature gradient was much
∗ Corresponding author at: University Lyon 1 – LAGEP – La Doua Campus –
lower.
3 rue Victor Grignard, Villeurbanne F-69616, France. Tel.: +33 4 72 43 18 70;
Tsotsas [7] conducted several experiments of vacuum con-
fax: +33 4 72 43 16 99. tact drying of hygroscopic packed beds. Contrary to Moyne, this
E-mail address: peczalski@lagep.univ-lyon1.fr (R. Peczalski). author did not observe a constant rate period and the falling rate

0255-2701/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2006.12.009
A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730 723

period started as soon as the drying began. During Tsotsas’s


experiments, all moisture was located in the particles (since the
particles were porous) and consequently, the liquid water trans-
fer was not possible from the bulk to the heating wall. Thus,
in this particular case, Tsotsas [7] succeeded in simulating the
experiment by using a classical quasi-stationary “vaporization
front” model.
The most recent data were published by Kohout et al. [8,9].
These authors developed two models in order to simulate the
drying rates of non-porous glass beads packed in a vessel of
cylindrical geometry: a general dynamic model and a specific
steady state model. The second was valid in the case where the
mass transfer of the liquid phase to the heating wall by capillary
flow was faster than the water vapour mass transfer from the
wall during the constant rate period. According to their second
model, during the constant rate period, the evaporation phenom-
ena occurred essentially in a thin zone near the heating wall and
the drying rates were governed both by heat transfer at this heat-
ing wall and by water vapour mass transfer through the bulk of
the granular bed. The average moisture content decreased from Fig. 1. Laboratory vacuum contact dryer.
its initial value to the percolation threshold value when redistri-
bution of the liquid by capillary flow stopped and the falling rate The vessel was 200 mm wide and 300 mm high. The bottom
period started. Then, a sharp drying front moved from the heat- plate of this vessel was made of stainless steel and supported
ing wall to the bulk of the bed which was divided into two zones: a temperature probe in contact with the first layer of powders.
a drying zone between the heated wall and the drying front and a The vessel side wall was made of glass. The vessel cover was
wet zone between the drying front and the free surface of the bed. made of stainless steel and supported a pressure probe and a
In the wet zone, the solvent content was supposed to be uniform temperature probe in contact with the bulk of the powder bed.
and equal to the percolation threshold as in a classical “vapor- The solvent vapour evaporated in the vessel was condensed
ization front” drying model. Whereas Kohout et al. [9] observed by means of a cooling fluid (methanol) circulating through the
good agreement between the experimental and simulated overall condenser (5). The residual solvent vapour was condensed inside
drying times, the simulated drying curves did not represent the a trap cooled with liquid nitrogen (6).
falling rate period of their experimental drying kinetics well. The total mass of the condensate, the inlet and outlet jacket
Nastaj [10] used a similar “vaporization front” model in order temperatures and the bulk and bottom product temperatures were
to simulate the falling rate period during the vacuum contact continuously recorded throughout the duration of the experi-
drying in packed beds of selected biotechnology products. A ments.
good agreement between simulations and experiments was only
observed in the case where the bed depth was small (2 mm). 2.2. Materials
According to our results, the classical “vaporization front”
model did not adequately represent the vacuum contact drying Two testing powdered materials, namely potassium chloride
data of monohydrate lactose and potassium chloride in packed and monohydrate lactose were selected, according to the follow-
beds during the falling rate period. As a result, modification ing criteria:
of the classical model was proposed resulting in a satisfactory
representation of our experimental data. • morphology close to that of industrial pharmaceutical pow-
ders;
2. Materials and methods • weakly hygroscopic;
• well known physical properties;
2.1. Apparatus • no major chemical and physical transformations during dry-
ing;
The experiments were conducted with a vacuum contact dryer
designed and set up in our laboratory as schematized in Fig. 1. Their properties are gathered in Table 1.
The main part (1) of the apparatus was constituted by an The solvent chosen was ethanol. Its properties are presented
8 liters jacketed cylindrical vessel heated from the bottom by in Table 2.
thermo fluid circulation (silicon oil). The fluid temperature was
controlled by a thermostat (2) in the range between 30 and 90 ◦ C. 2.3. Drying experiments protocol
The pressure inside the vessel was controlled by means of a
piezoelectric pressure probe, a vacuum pump (3) and an electro The initial and final (after drying) product solvent content
valve (4) in the range of 1500–10,000 Pa. values were determined by weighing three samples before and
724 A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730

Table 1
Physical properties of potassium chloride and monohydrate lactose
Physical properties Potassium chloride Monohydrate lactose

Particle mean diameter, dpa (m) 467 × 10−6 47 × 10−6


Thermal conductivity of dry layer, λd (W m−1 K−1 ) 0.169 0.056 (aerated), 0.074 (tapped)
Density of dry layer, ρd (kg m−3 ) 1086 695 (aerated), 864 (tapped)
Porosity of dry layer, ψd 0.45 0.55 (aerated), 0.44 (tapped)

Table 2 3. Modeling
Physical properties of ethanol at 25 ◦ C
Physical properties Ethanol 3.1. The classical “vaporization front” model
Specific heat of liquid, Cpl (J kg−1 K−1 ) 2840
Specific heat of vapour, Cpv (J kg−1 K−1 ) 1424 The steady state model proposed by Kohout et al. [8] was used
Latent heat of evaporation, Hvap (J kg−1 ) 9.186 × 10−5 to simulate both the constant rate and the falling rate periods.
Absolute viscosity of vapour, ηv (kg m−1 s−1 ) 8.6 × 10−6
Thermal conductivity of vapour, ␭v (W m−1 K−1 ) 0.0145
3.1.1. The constant rate period
Surface tension: σ (N m−1 ) 0.0224
Density of liquid, ρl (kg m−3 ) 790 During this period, the evaporation was supposed to occur
at the heating wall and the drying rate was determined by the
simultaneous solution of steady state heat and mass transfer
Table 3 equations.
Operating conditions of drying runs with monohydrate lactose The vapour flux from the wall to the bed surface was described
Operating conditions Trial 1 Trial 2 Trial 3
by Darcy’s law which was given by the relationship expressed
with the formula:
Temperature of hot fluid, Tf (◦ C) 56.3 57.2 57.2  
Operating total pressure, Pvessel (Pa) 7000 7000 7000 Kv ρv Pvessel − Psat,wall (Tsat,wall )
ṁ = − (1)
Initial solvent content, Xini (%) 48 17 15 ηv zmax
where Kv represents the packed bed permeability of vapour,
ρv the density of vapour, zmax the depth of the bed, Psat,wall
after complete dehydratation in an oven at 110 ◦ C during 1 day the saturation pressure at the heating wall and Tsat,wall is the
for potassium chloride runs and at 80 ◦ C during 1 day for mono- saturation temperature at the heating wall.
hydrate lactose runs. The heat flux at the heating wall was expressed by the clas-
The drying rate curve was obtained by derivating the solvent sical relationship:
content versus time curve smoothed by a five points running
average. The last one was determined from the condensate mass q̇ws = hf (Tf − Tsat,wall ) (2)
recording.
The heat transfer coefficient of the jacketed vessel (hf ) was In steady state conditions, the heat flux was entirely consumed
experimentally determined by evaporating experiments of pure by water vaporization:
water inside the dryer. During water boiling, the vapour flow rate q̇ws
(which represents the heat flux), the hot fluid temperature and ṁ = (3)
Hvap
the internal vessel surface temperature were recorded and their
steady state values were used for calculating the hf coefficient The saturation pressure and temperature are related by the
values (hf = 200 W m−2 K−1 ). following equilibrium equation (Antoine’s equation):
For monohydrate lactose, three experiments were conducted. 3 /T )+53.17+5.95×10−7 T 2
During one of these experiments, several samples were removed Psat = 103 e−5.09 log(Tsat )−(6.61×10 sat sat (4)
from the upper part of the bed at different drying times. The By combining the four above equations, we could calculate
operating conditions of these are given in Table 3. the saturation temperature at the heating wall and the solvent
For potassium chloride, three experiments were realized. The vapour or the heat flux.
operating conditions of these trials are given in Table 4. The average solvent content decreased from the initial con-
ditions to a critical solvent content (Xcr ) at which the falling
Table 4
rate period begins. The evolution of mean solvent content of the
Operating conditions of drying runs with potassium chloride powder, noted X, with time was given by the relationship:
Operating conditions Trial 4 Trial 5 Trial 6 dX ṁA
= (5)
Temperature of hot fluid, Tf (◦ C) 55.6 56.9 67.6 dt md
Operating total pressure, Pvessel (Pa) 7000 7000 7000
Initial solvent content, Xini (%) 34 10 33
where A represents the heating wall area and md is the dry
product mass.
A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730 725

The bed permeability of the solvent vapour and the critical By combining Eqs. (7) and (8), the vapour flux could be
solvent content were obtained by fitting the calculated curves to calculated by the following expression:
the experimental ones.
Tf − Tsat,front
ṁ = (9)
Hvap ((1/ hf ) + (zfront /λd ))
3.1.2. The falling rate period
During this period a drying front moved from the heating The saturation pressure and temperature were still related by
wall to the bulk of the bed. In the dry layer, the heat conduc- the classical Antoine’s Eq. (4). As well, the solvent balance at
tion mechanism was mainly by conduction. In the wet layer, the the vaporization front was expressed by
solvent vapour mass transfer was supposed to take place by per- dzfront
meation. During the entire falling rate period, the solvent content ṁ = ρd Xfront (10)
dt
at the front (Xfront ) was supposed to be equal to the critical sol-
vent content. Except for the mass transfer equation, the model where Xfront represents the solvent content in the wet zone near
presented below proceeds from the well known phase change the vaporization front.
front approach developed by Plank (Grigull and Sandner [11] The evolution of solvent content with time was still described
and other ‘Heat Conduction’ textbooks). The scheme of tem- by Eq. (5).
perature and solvent content profiles through the bed during this By combining Eqs. (9) and (10), the vaporization front rate
period is presented in Fig. 2. could be calculated by the following equation:
The vapour flux in the wet zone above the drying front was  
dzfront (Tf − Tsat,front )
expressed by the following Darcy’s law: = (11)
dt ρd × Xfront ×Hvap × (1/ hf + zfront /λd )
 
Kv ρv Pvessel − Psat,front (Tsat, front ) Furthermore, by combining Eqs. (4), (6), (9) and (11), we
ṁ = − (6)
ηv zmax − zfront could calculate the saturation temperature and the solvent vapour
or the heat flux at the vaporization front.
where Psat,front represents the saturation pressure at the vaporiza- During this period, as soon as the vaporization front pene-
tion front, Tsat,front the saturation temperature at the vaporization trated just a few millimetres through the bed, the drying rate
front and zfront is the distance from the heating wall to the vapor- was mainly controlled by the heat conduction into the dry layer
ization front. and the solvent vapour permeation through the wet zone could
Moreover, the heat flux at the vaporization front was given be neglected. Thus, Eq. (6) could be removed and Eqs. (7)–(11)
by could be used by replacing Tsat,front by Tsat,vessel (Pvessel ).
 
Tf − Tsat,front 3.2. The new “vaporization front with a varying solvent
q̇front = (7)
(1/ hf ) + (zfront /λd ) supply” model

The thermal energy balance at the vaporization front was 3.2.1. The constant rate period
expressed by the following steady state relationship: To simulate this period, Eqs. (1)–(5) of the classical model
were used. During this period, the average solvent content
q̇front
ṁ = (8) decreased from the initial conditions to the first critical solvent
Hvap content, noted Xcr1 . The permeability of the bed and first criti-
cal solvent content values were obtained by fitting the simulated
curves to the experimental data.

3.2.2. The “transition” period


This period corresponded to the transition between the con-
stant rate period and the starting of the vaporization front
progression. During this period, the amount of liquid at the heat-
ing wall decreased but the vaporization still occurred only in a
thin zone near the heating wall. As a result, the equations cor-
responding to the constant rate period were used. Nevertheless,
for bed solvent contents ranging from the first critical solvent
content to the second critical value, noted Xcr2 , the heat transfer
coefficient between the hot fluid and the first layer of particles,
noted hg,tran , was supposed to vary linearly with the average
solvent content as expressed by the following relationships:
 
hf − h g
Fig. 2. Scheme of temperature and solvent content profiles through the bed hg,tran = (X − Xcr2 ) + hg (12)
during the falling rate period of the classical model. Xcr1 − Xcr2
726 A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730

with: determined from the following equation:


 −1  Xfin  Zmax
1 1 −md dX = Aρd (Xfront − Xfin ) dz (15)
hg = + (13)
hf hws Xcr2 0

where Xfin represents the final solvent content and which led
The contact heat transfer coefficient (hws ) was calculated by
after integrating to the relationship:
the method proposed by Schlünder [12] and Mollekopf and Mar-
tin [13] and described in details by Schlünder and Mollekopf [1]. (aXcr2 − Xfin + b) e−a − ((a − 1)Xfin + b) = 0 (16)
The values of the parameters necessary to apply this correlation
can be found in Tsotsas [7]. As for the classical “vaporization front” model, from the early
Besides, the q̇ws , ṁ and dX/dt values were calculated in the stage of the falling rate period, the drying rate was controlled
same manner than during the constant rate period by replacing by the heat flux through the dry layer. It is worth noting that the
the parameter hf by the hg,tran expression given by Eq. (12). key parameters of this last period modeling were the dry layer
The second critical solvent content was also obtained by thermal conductivity and the variation of the solvent content at
fitting the simulated curves to the experimental data. the vaporization front expressed by Eq. (14).
If the drying period started at an initial solvent content lower
than Xcr2 , which is generally the case when the drying was
3.2.3. The falling rate period
operated after a filtration step, then, in this case, only the sol-
During this period, a drying front was supposed to move from
vent supply parameter (a) was fitted instead of the four fitting
the heating wall to the bulk of the bed. The previously described
parameters (a, Kv , Xcr1 and Xcr2 ) for the general case.
classical model was modified by introducing a linear variation
of the solvent content at the vaporization front with the bed
4. Results and discussion
average solvent content. Thus, at the beginning of the falling
rate period, the solvent content at the vaporization front was
4.1. Simulations with the classical “vaporization front”
higher than the second critical solvent content so that the drying
model
rate was increased. In the same way, at the end of the falling rate
period, the solvent content at the vaporization front was lower
The simulated (with the classical model) curves corre-
than the second critical solvent content so that the drying rate
sponding to experimental conditions of runs 1 and 4 (see
was decreased. The scheme of temperature and solvent content
Tables 3 and 4) are presented in Figs. 4 and 5 and range over the
profiles through the bed during this period is presented in Fig. 3.
constant rate and the falling drying rate periods. During trial 1,
Then, Eqs. (6)–(11) were applied by replacing hf by hg .
the bed was much more compact than for trial 2 due to the much
A new fitting parameter, called the solvent supply parame-
higher initial solvent content and consequently, the “tapped”
ter, noted a, was used to express the vaporization front solvent
thermophysical properties (see Table 1) of monohydrate lactose
content at follows:
were used to simulate the trial 1. For trial 2, the “aerated”
Xfront = aX + b (14) thermophysical properties (see Table 1) of monohydrate lactose
were used. The simulated (with the classical model) curves
In this last equation, the parameter b is an empirical parameter corresponding to experimental conditions of runs 2 and 5 (see
which enabled to have a solvent content equal to zero when the Tables 3 and 4) are presented in Figs. 4 and 5 and range only
drying front reached the surface of the bed. Next the b value was over the falling rate period. In the case of potassium chloride, the
bed had practically the same (“aerated”) density whatever was

Fig. 3. Scheme of temperature and solvent content profiles through the bed Fig. 4. Measured and simulated (with the classical model) drying rate curves:
during the falling rate period of the new model. monohydrate lactose (trials 1 and 2).
A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730 727

Fig. 5. Measured and simulated (with the classical model) drying rate curves: Fig. 6. Solvent content as a function of time measured by sampling the upper
potassium chloride (trials 4 and 5). part of the bed and by “condensate” mass recording during trial 3.

the initial solvent content. Identified values of vapour perma- front exists inside the bed since its lower part dried more quickly
bilities and critical solvent contents for each trial are provided than its upper part.
in Table 5. From the above observations, we can conclude the classi-
The small disturbance on the drying rate curve for trial 1 cal model did not precisely predict the drying curves during
(see Fig. 1) at a solvent content of about 8% was due to an the falling rate period of vacuum contact drying of crystalline
experimental malfunction. powder packed beds.
Concerning the constant rate period of trials 1 and 4, the only As a matter of fact, Moyne [5] showed that the first falling
drying rate controlling parameter is the packed bed permeability rate period can be represented by the classical model but not the
of vapour (Kv ). This parameter was adjusted so a necessarily second one since at the end of the drying, a drying rate not equal
good agreement between measured and simulated drying rate to zero was obtained despite the fact that the solvent content in
curves was obtained. the wet zone remained constant.
Concerning the falling rate period, the drying rate curves Moreover, Schlünder [14] criticized the classical “vaporiza-
simulated with the classical model merged one to another shortly tion front” model generally used to describe the falling rate
after the critical solvent content. The experimental curves of period of porous material during convective drying. According
trials 1 and 4, starting at high initial solvent contents, are located, to this author, application of this model resulted in considerable
respectively, above the experimental curves of trials 2 and 5, temperature gradients within the dry zone which actually did
starting at low solvent contents. Moreover, the simulated curves not exist during convective drying. Then, Schlünder developed
for trials 2 and 5 are significantly located below the experimental a new model called “wet surface model” based on the assump-
ones so that the classical model underestimates the experiments. tion that the dry and wet zones are arranged in “parallel” and
Furthermore, the solvent content evolutions as a function not in “series”, with the solvent being transported to the heating
of time measured by sampling the upper part of the bed and surface by capillary flow.
by “condensate” mass recording during trial 3 are presented in In the case of vacuum contact drying, strong temperature
Fig. 6. The initial solvent content of this trial corresponded to gradients were experimentally observed near the heating wall
the beginning of the falling rate period. by Moyne [5] and Baillon [6] which indicates that a vaporiza-
The solvent content of the samples extracted from the upper tion front exists and moves through the bed. Nevertheless, we
part of the bed decreased as soon as the drying started. This believe that Schlünder’s approach [14] is really interesting as it
observation is inconsistent with the classical “vaporization introduced a liquid water transport to the vaporization front even
front” model hypothesis, in which the solvent content of the during the falling rate period.
wet zone above the vaporization front should be constant during Consequently, the basic hypothesis of our new modeling was
the entire duration of the falling rate period. to suppose that the vaporization front was supplied with sol-
Nevertheless, the solvent content in the upper part of the bed vent by gravitational permeation or capillary diffusion during
decreased more slowly than the average solvent content mea- the falling rate period. In the frame of a ‘vaporization front’
sured by condensate which seemed to indicate that a vaporization model, this assumption implied an artificial front solvent content

Table 5
Identified parameters of drying runs (classical “vaporization front” model)
Identified parameters Trial 1 Trial 2 Trial 4 Trial 5

Packed bed permeability of vapour, Kv (m2 ) 4.3 × 10−13 4.3 × 10−13 1.2 × 10−12 1.2 × 10−12
Critical solvent content, Xcr 15.0% – 10.0% –
728 A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730

Table 6
Identified parameters of drying runs (new “vaporization front with a varying solvent supply” model)
Identified parameters Trial 1 Trial 2 Trial 4 Trial 5

Packed bed permeability of vapour, Kv (m2 ) 4.3 × 10−13 4.3 × 10−13 1.2 × 10−12 1.2 × 10−12
First critical solvent content, Xcr1 21.0% – 16.0% –
Second critical solvent content, Xcr2 17.5% – 11.0% –
Solvent supply parameters, a 10 1.7 3.5 1

higher than the second critical solvent content and thus resulted
in enhanced drying rates. It must be emphasized that the empir-
ical ‘solvent supply parameter’ (a) was intended to represent
a liquid solvent flux arriving at the front and that the result-
ing front solvent content increase was an artefact of the simple
‘front’ approach. In reality, there was no a sharp line delimiting
the wet and dry zone and there was a continuous liquid solvent
content gradient from a maximum at the top and the minimum at
the bottom of the bed. In our approach, the artificial front solvent
content overshoot corresponded to the solvent that moved from
the wet to the dry zone and was vaporized in the dry zone.
In order to discriminate between gravitational and capillary
flow within the wet zone (directed toward the vaporization front)
the Bond number was calculated. This dimensionless number is
Fig. 7. Experimental and simulated (with the new model) drying rates: mono-
the ratio of gravitational potential to capillary potential and is hydrate lactose (trials 1 and 2).
expressed by
ρl g
Bo = (17)
σ/r 2
where ρl is the density of liquid solvent, g the acceleration due
to gravity, σ the surface tension and r is the mean void radius of
the bed. This radius was considered equal to the mean crys-
tal radius in our calculation (see Table 2). The Bo value of
1.9 × 10−4 was obtained for the fine grained lactose monohy-
drate and 1.9 × 10−2 for the coarse grained potassium chloride.
Both values indicated that the capillarity dominated strongly
over gravity.

4.2. Simulations with the new “vaporization front with a


varying solvent supply” model
Fig. 8. Experimental and simulated (with the new model) drying rates: potas-
sium chloride (trials 4 and 5).
The simulated (with the new model) curves corresponding
to experimental conditions of runs 1 and 4 (see Tables 3 and 4)
over, the data of Table 7 showed that the overall drying times
with the new model are presented in Figs. 7 and 8 and range
needed to achieve a fixed final solvent content equal to 0.5%
over the constant rate and falling drying rate periods. The simu-
were also well predicted by this new model.
lated (with the new model) curves corresponding to experimental
One might observe that the difference between the exper-
conditions of runs 2 and 5 (see Tables 3 and 4) are presented in
imental drying curves of trials 1 and 2 corresponding to
Figs. 7 and 8 and range only over the falling rate period. For
trial 1, the “tapped” thermophysical properties (see Table 1) of
monohydrate lactose were used whereas for trial 2 the “aerated” Table 7
Predicted and experimental drying times for final solvent content equal to 0.5%
thermophysical properties (see Table 1) were used. For trials 4 (trials 1, 2, 4 and 5)
and 5, the “aerated” density of potassium chloride was used.
Identified values of vapour permeabilities, the first and second Trials Predicted drying times (min) Experimental drying
times (min)
critical solvent contents and solvent supply parameters for each
trial are provided in Table 6. 1 161.3 (from 25%) 171.2 (from 25%)
As shown in Figs. 7 and 8, a very good agreement between 2 470.0 (from 16%) 471.6 (from 16%)
4 99.6 (from 25%) 101.4 (from 25%)
experimental and simulated (with the new model) curves was 5 96.7 (from 8%) 101.7 (from 8%)
obtained over the entire duration of the drying process. More-
A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730 729

for the falling rate period did not precisely represent the exper-
imental drying rates.
Consequently, a new “vaporization front with a varying sol-
vent supply” model was developed. Primary model improvement
consisted of assuming linear variation of the solvent content at
the vaporization front, contrary to classical modeling for which
this parameter was assumed constant. This new model required
only one fitting parameter if the drying process started at such
a low solvent content that the falling rate period had already
begun.
To validate this new model, experiments were conducted
with a laboratory vacuum contact dryer. Drying rates of
monohydrate lactose and potassium chloride for different
heating wall temperatures were measured. Good agree-
Fig. 9. Experimental and simulated (with the new model) drying rates: potas-
ment between the simulated and experimental curves was
sium chloride (trial 6).
observed and the experimental drying times were accurately
predicted.
monohydrate lactose was much higher than the difference Moreover, the final objective of this study was to simulate
between the experimental drying curves of trials 4 and 5 cor- the drying rates during vacuum contact drying with intermit-
responding to potassium chloride. Concerning the simulations, tent stirring using the new model developed in this work to
this effect turned into an identified ‘solvent supply parame- simulate the period without stirring. With this new approach,
ter’ (representing the mobility of the solvent) much lower for we will be able to determine, on a more scientific basis than
potassium chloride than for lactose monohydrate. These results the usual “trial and error” method, the optimal intermittent
corroborated with the Bond number values indicating that the stirring sequence parameters (duration, frequency) which cor-
mobility of ethanol by capillary flow within the powder bed respond to the best compromise between the drying times
is much higher for monohydrate lactose (Bo = 1.9 × 10−4 ) than and attrition defects that usually affects the quality of dried
for potassium chloride (Bo = 1.9 × 10−2 ). Another explanation powder.
for the difference between the drying trials was that the liquid
solvent flux at the vaporization front depended on the solvent
content of the wet zone and thus that the “solvent supply param- Acknowledgements
eter” (a) was probably a function of the second critical solvent
content (Xcr2 ). According to Table 6, these two parameters This work was financially supported by the French Agency
appear effectively correlated. for Environment and Energy Management ‘ADEME’ and the
pharmaceutical group ‘SANOFI-AVENTIS’.
4.3. Validation of the new “vaporization front with a
varying solvent supply” model Appendix A. Nomenclature

Moreover, our new model was validated by conducting one


experiment and the corresponding simulation with other oper- a solvent supply parameter
ating conditions that the ones used for identification of the A heating wall area (m2 )
parameters optimal values. b empirical parameter (Eqs. (14) and (16))
The simulated drying rate data corresponding to trial 6 pre- Bo Bond number
sented in Fig. 9 have been obtained by using the same values Cp specific heat (J kg−1 K−1 )
of the identified parameters (Kv , Xcr1 , Xcr2 and a) than the ones dpa particle mean diameter (m)
optimised with the data for trial 4. g acceleration due to gravity (m2 s−1 )
The experimental and simulated drying times needed to hf fluid-wall heat transfer coefficient (W m−2 K−1 )
obtain a final solvent content equal to 0.5% starting from 23% hg overall heat transfer coefficient (W m−2 K−1 )
are equal to 71.5 and 67.9 min, respectively. hws contact heat transfer coefficient (W m−2 K−1 )
Consequently, as shown in Fig. 9, good agreement between H latent heat (J kg−1 )
simulated and experimental drying rates was obtained and the K packed bed permeability (m2 )
overall drying time was also accurately predicted, with our new ṁ drying rate (kg m−2 s−1 )
model. md dry product mass (kg)
P pressure (Pa)
5. Conclusions Pvessel operating total pressure (Pa)
q̇front heat flux at the vaporization front (W m−2 )
It was confirmed that the vacuum contact drying model of q̇ws heat flux at the heating wall (W m−2 )
packed beds based on the classical “vaporization front” approach r mean void radius of the bed (m)
730 A. Michaud et al. / Chemical Engineering and Processing 47 (2008) 722–730

t time (s) References


T temperature (K)
X solvent content (kg solvent kg−1 dry product) [1] E.U. Schlünder, N. Mollekopf, Vacuum contact drying of free flowing
mechanically agitated particulate material, Chem. Eng. Process. 18 (1984)
Xcr1 first critical mean solvent content (kg solvent kg−1 dry 93–111.
product) [2] E. Tsotsas, E.U. Schlünder, Vacuum contact drying of free flowing mechan-
Xcr2 second critical mean solvent content (kg solvent kg−1 ically agitated particulate multigranular packing, Chem. Eng. Process. 20
dry product) (1986) 339–349.
z distance from the heating wall (m) [3] E. Tsotsas, E.U. Schlünder, Vacuum contact drying of mechanically agi-
tated beds: the influence of hygroscopic behaviour on the drying rate curve,
zmax depth of the bed (m) Chem. Eng. Process. 21 (1987) 199–208.
[4] D. Farges, M. Hemati, C. Laguérie, F. Vachet, P. Rousseaux, A new
Greek letters approach to contact drying modeling, Dry. Technol. 13 (1995) 1317–1329.
η absolute viscosity (kg m−1 s−1 ) [5] C. Moyne, Transferts couples chaleur—masse lors du séchage: prise en
λ thermal conductivity (W m−1 K−1 ) compte du mouvement de la phase gazeuse, Thèse, Institut National Poly-
technique de Lorraine, 1987.
ρ density (kg m−3 ) [6] B. Baillon, Séchage sous vide micro-ondes combiné de granulés pharma-
σ surface tension (N m−1 ) ceutiques, Thèse, Université de Pau et des Pays de l’Adour, 1996.
ψ porosity [7] E. Tsotsas, Über den Einfluß der Dispersität und der Hygroskopizität auf
den Trocknungsverlauf bei der Vakuum-Kontakttrocknung rieselfähiger
Subscripts Trocknungsgüter, Dissertation, Universität Karlsruhe, 1985.
[8] M. Kohout, A.P. Collier, F. Stepanek, Mathematical modelling of solvent
b bulk drying from a static particle bed, Chem. Eng. Sci. 61 (2006) 3674–3685.
cr critical [9] M. Kohout, A.P. Collier, F. Stepanek, Multi-scale analysis of vacuum con-
d dry tact drying, in: Proceedings of the IDS 2006 B, 2006, pp. 1087–1094.
f hot fluid [10] J.F. Nastaj, Vacuum contact drying of selected biotechnology products,
fin final Dry. Technol. 12 (1994) 1145–1166.
[11] U. Grigull, H. Sandner, in: B. Munz Brienza (Ed.), Heat Conduction,
front vaporization front Hemisphere Publishing Corporation, United States of America, 1984.
ini initial [12] E.U. Schlünder, Wärmeübergang an bewegte Kugelshüttungen bei
l liquid kurfristigem Kontakt, Chem.-Ing.-Technol. 43 (1971) 651–654.
sat saturation [13] N. Mollekopf, H. Martin, Zur Theorie des Wärmeübergangs an bewegte
tran transition period Kugelschüttungen bei kurzfristigem Kontakt, VT-Verfahrenstechnik 16
(1982) 701–706.
v vapour [14] E.U. Schlünder, Drying of porous material during the constant and the
vap evaporation falling rate period: a critical review of existing hypotheses, Dry. Technol.
wall heating wall 22 (2004) 1517–1532.

You might also like