This article was downloaded by: [University of Connecticut]

On: 12 October 2014, At: 07:20

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

Experimental Study and Modeling of Crystalline

Powders Vacuum Contact Drying with Intermittent
Stirring
a a a
A. Michaud , R. Peczalski & J. Andrieu
a
Université de Lyon, Laboratoire d'Automatique et de Génie des Procédés—LAGEP ,
Villeurbanne, France
Published online: 30 Jul 2007.

To cite this article: A. Michaud , R. Peczalski & J. Andrieu (2007) Experimental Study and Modeling of Crystalline Powders
Vacuum Contact Drying with Intermittent Stirring, Drying Technology: An International Journal, 25:7-8, 1163-1173, DOI:
10.1080/07373930701438501

To link to this article: http://dx.doi.org/10.1080/07373930701438501

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
 Drying Technology, 25: 1163–1173, 2007
Copyright # 2007 Taylor & Francis Group, LLC
ISSN: 0737-3937 print/1532-2300 online
DOI: 10.1080/07373930701438501

Experimental Study and Modeling of Crystalline Powders

Vacuum Contact Drying with Intermittent Stirring
A. Michaud, R. Peczalski, and J. Andrieu
Université de Lyon, Laboratoire d’Automatique et de Génie des Procédés—LAGEP,
Villeurbanne, France

determination of optimal stirring and drying parameters

on a rational basis.
Downloaded by [University of Connecticut] at 07:20 12 October 2014

Modified literature models were successfully applied to simulate

the evolution of the average solvent content (ethanol) of potassium The vacuum contact drying of powder stirred bed has
chloride bed during vacuum contact drying with intermittent been investigated in detail by Schlünder and Mollekopf [1]
stirring. Our new modeling approach incorporated the following for a non-hygroscopic free-flowing granular material
modifications: the introduction of a jacketed vessel heat transfer
coefficient and an accumulation term for the heating wall tempera- and extended by other authors to more complex cases,
ture; the application of alternate static bed and stirred bed con- including:
ditions; and modeling of all the three drying phases, namely the
constant rate phase, the transition phase, and the falling rate phase.
poly-dispersed materials;[2]
Moreover, several validation experiments were carried out with dif-
hygroscopic materials;[3,4]
ferent operating conditions to identify the values of the unknown
materials with a binary mixture;[5]
model parameter. The optimal stirring conditions were investigated
modeling with particle intrinsic water transfer
by calculating total drying times for different sequences of stirring
and no stirring periods. resistance;[6]

modeling with sticking particles and crust
Keywords Crystalline powders; Experimental drying kinetics; formation at heating wall.[7]
Modeling; Packed bed; Stirred bed; Vacuum contact
drying As concerns vacuum contact drying of powder packed
bed, some experimental data and theoretical analysis have
been published by Tsotsas,[8] Moyne,[9] Baillon,[10] and
INTRODUCTION more recently by Kohout et al.[11,12] Kohout[11,12]
developed two models in order to simulate the drying rates
Vacuum contact heating is widely used in the pharma-
of non-porous glass beads packed in a cylindrical vessel: a
ceutical industry to dry granular products that are sensitive
general dynamic model and a specific steady-state model.
to oxygen and temperature and are often toxic and
The steady-state model was valid in the case where the
explosive.
liquid mass transfer to the heating wall by capillary flow
In the pharmaceutical industry, the major challenge dur-
was faster than the vapor mass transfer from the wall
ing powder drying is the preservation of the initial grain
during the constant rate phase. According to this model,
size and shape during the process by preventing the
during the constant rate phase, the vaporization phenom-
agglomeration or=and attrition phenomena. Regarding
ena occurred essentially in a thin zone near the heating wall
attrition, which is primarily due to the mechanical impact
and the drying rates were governed both by contact heat
of the stirrer blades on the individual grains, the simplest
transfer at this heating wall and by solvent vapor mass
solution is to reduce the stirring rate or the stirring dur-
transfer through the bulk of the bed. When the redistri-
ation. This is why the influence of intermittent stirring,
bution of the liquid solvent by capillary flow stopped and
which alternates stirring periods and no stirring periods,
the average solvent content decreased to its percolation
on the drying rate was experimentally investigated in
threshold value, the falling rate phase was supposed to
this work and then modeled, which allowed for the
start.
During the falling rate phase, a sharp vaporization front
Correspondence: R. Peczalski, Université de Lyon, Université
Lyon 1, ESCPE Lyon, CNRS, UMR 5007, Laboratoire d’Auto- moved from the heating wall through the bulk of the bed,
matique et de Génie des Procédés—LAGEP, Villeurbanne which then could be divided into two zones: a dry zone
F-69616, France; E-mail: [email protected] between the heating wall and the drying front and a wet

1163
 1164 MICHAUD ET AL.

zone between the drying front and the free surface of the and supported a temperature probe in contact with the first
bed. In the wet zone, the solvent content was supposed to layer of the powder bed at the vessel wall. The vessel side
be uniform with a value equal to the percolation threshold wall was made of glass. The vessel cover was made of stain-
as in a classical ‘‘vaporization front’’ drying model. less steel and supported a pressure gauge and a temperature
Even if the theoretical basis of the vacuum contact dry- probe in contact with the bulk of the powder bed.
ing with continuous stirring and without stirring are now The powder was mechanically agitated by a two-blade
well established and available in the literature, no experi- stirrer driven by an electric motor (5). The stirrer rotational
mental data or corresponding modeling and simulations speed could be set in the range of 23 to 133 rpm. A timer
for vacuum contact drying with intermittent stirring have was installed in order to set up the sequences and the
been previously published. duration of stirring and no stirring periods.
The solvent vapor flowing out the vessel was condensed
EXPERIMENTS
by means of a tubular condenser cooled with methanol (6).
Apparatus The residual solvent vapor was trapped inside a recipient
The experiments were conducted with a vacuum contact cooled with liquid nitrogen (7).
dryer designed and set up in our laboratory as schematized The total mass of the solvent condensate, the inlet and
on Fig. 1. outlet heating fluid temperatures, and the bulk and bottom
Downloaded by [University of Connecticut] at 07:20 12 October 2014

The main part (1) of the laboratory dryer was constituted powder temperatures were continuously recorded through-
by an 8-L jacketed cylindrical vessel heated by thermo fluid out the duration of the contact drying experiments.
circulation (silicon oil). The fluid temperature was con-
trolled by a thermostat (2) in the range between 30 and
Materials
90C. The pressure inside the vessel was controlled by means
Potassium chloride was selected as a testing material
of a piezoelectric pressure probe, a vacuum pump (3), and an
according to the following criteria:
electro valve (4) in the range between 1500 and 10,000 Pa.
The vessel was 200 mm wide and 300 mm high. The
morphology close to that of some industrial phar-
bottom plate of this vessel was made of stainless steel maceutical powders;

FIG. 1. Laboratory vacuum contact dryer with stirring.

 EXPERIMENTAL STUDY OF CRYSTALLINE POWDERS 1165

TABLE 1 and the initial solvent content was around 33% kg of

Physical properties of potassium chloride ethanol=kg of dry product.
Physical properties Potassium chloride
MODELING OF VACUUM CONTACT DRYING
Particle mean diameter: dpa (m) 0.467  10 3
WITH INTERMITTENT STIRRING
Thermal conductivity of dry layer: 0.169
According to these experimental drying rate curves,
kd (W m 1 K 1)
three distinct drying phases could be identified:
Density of dry layer: qd (kg m 3) 1086
Porosity of dry layer: Wd 0.45
a classical constant rate phase during which the
powder bed at the heating wall remained saturated
with solvent;

weakly hygroscopic;

a ‘‘transition’’ phase, starting from a critical sol-

well-known physical properties;
vent content practically independent of operating

no major chemical and physical transformations
conditions, during which the drying rates started
during contact drying experiments.
to decrease slowly;
Its main properties are gathered in Table 1. The solvent
a falling rate phase, starting from a critical solvent con-
Downloaded by [University of Connecticut] at 07:20 12 October 2014

was pure ethanol. Its properties are presented in Table 2. tent practically independent of operating conditions,
during which the drying rate decreased strongly.
The Constant Rate Phase
Drying Experiments Protocol
Stirring ‘‘Off’’ Periods
The initial and final (after drying) solvent content values
The steady-state model previously proposed by
of the product were determined by weighing three samples
Kohout[11] was used. In this modeling, the vaporization
before and after complete dehydration in an oven at 110C
was supposed to occur at the heating wall and the drying
during one day.
rate was calculated by solving simultaneously and numeri-
The drying rate curve was obtained by derivating the
cally the steady state heat and mass transfer equations.
solvent content versus time curve. This curve was determ-
The vapor flux from the wall to the bed surface was
ined from the condensate mass recording data which were
described by Darcy’s law expressed by the following
smoothed by a 5-point running average method.
relationship:
Moreover, the overall heat transfer coefficient value of the

jacketed vessel (hf) was experimentally determined by pure K q Pvessel  Psat;wall ðTsat;wall Þ
m_ ¼  v;ns v ð1Þ
water boiling experiments inside the dryer. During these boil- gv zmax
ing runs, the condensate mass flow rate (which represents the
heat flux), the heating fluid temperature, and the internal ves- where Kv,ns represents the vapor bed permeability during
sel surface temperature were recorded and their steady-state the stirring ‘‘off’’ periods, qv the density of vapor,
values were used for calculating the coefficient values. zmax the depth of the bed, Psat,wall the saturation pressure
Five vacuum contact drying experiments were realized. of the solvent at the heating wall, and Tsat,wall the satu-
The operating conditions of these trials are given in ration temperature at the heating wall.
Table 3. The corresponding drying rate curves are The heat flux at the heating wall (q_ ws ) was expressed by
presented and analyzed in Results and Discussion. the classical relationship:
For each trial, the mass of dry product (md) was around q_ ws ¼ q_ f ¼ hf ðTf  Tsat;wall Þ ð2Þ
1.025 kg, the depth of the bed (zmax) was around 0.065 m,
In steady-state conditions, the heat flux was entirely used
for the solvent vaporization, which led to the classical
TABLE 2 relationship:
Thermophysical properties of ethanol at 25C
q_ ws
Physical properties Ethanol _ ¼
m ð3Þ
DHvap
1 1
Specific heat of vapor: Cpv (J kg K ) 1424
9.186  10 5 If the solvent is ethanol, the saturation pressure and
Latent heat of vaporization:
temperature were related by the following equilibrium
DHvap (J kg 1)
Absolute viscosity of vapor: 8.6  10 6 equation (Antoine’s equation):
gv (kg m 1 s 1)  
6:61:103 7 2
Thermal conductivity of vapor: 0.0145 Psat ¼ 103 e 5:09 lnðTsat Þ Tsat þ 53:17 þ 5:95:10 Tsat ð4Þ
kv (W m 1K 1)
with Tsat in K and Psat in Pa.
 1166 MICHAUD ET AL.

TABLE 3
Operating conditions of drying runs
Operating conditions Trial 1 Trial 2 Trial 3 Trial 4 Trial 5
Mean temperature of hot fluid: Tf (C) 46.4 47.2 46.8 39.4 38.6
Operating total pressure: Pvessel (Pa) 7000 7000 7000 7000 5000
Stirring rate: Z (rpm) 23 23 23 23 23
Stirring ‘‘on=off’’ time (min) continuous 2.5=5 5=5 2.5=5 2.5=5

Then, by combining the four above equations, we could heating fluid and the first layer of particles, noted hg,tran,
calculate the saturation temperature at the heating wall and was supposed to decrease linearly with the average solvent
the solvent vapor flux or the heat flux at the heating wall. content as expressed by the following relationships:
The average solvent content decreased from the initial

hf  hg
value to a first critical solvent content, noted Xcr1, at which hg;tran ¼ ðX  Xcr2 Þ þ hg ð6Þ
the transition phase started. The time evolution of the aver- Xcr1  Xcr2
Downloaded by [University of Connecticut] at 07:20 12 October 2014

age powder solvent content, noted X, was obtained by the where

following equation:

1
dX A 1 1
¼m
_  ð5Þ hg ¼ þ ð7Þ
dt md hf hws

where A represents the heating wall area in contact with the The contact heat transfer coefficient at the interface
product and md the total dry product mass. between the heating wall and the dry layer value, noted
The vapor bed permeability during the stirring ‘‘off’’ hws, was calculated by the method proposed by Schlünder[13]
periods and the first critical solvent content values were and Mollekopf and Martin[14] and described in detail by
obtained by fitting the calculated drying rate curves to Schlünder and Mollekopf [1] and Tsotsas.[8] Then, the q̇ws,
the experimental ones. ṁ, and dX=dt values were calculated in the same manner
than during the constant rate phase by replacing hf by hg,tran
given by Eq. (6).
Stirring ‘‘On’’ Periods The second critical solvent content was also obtained by
As like for stirring ‘‘off’’ periods, the solvent vaporiza- fitting the simulated curves to the experimental ones.
tion was supposed to occur at the heating wall and the
drying rate values were determined by the simultaneous Stirring ‘‘On’’ Periods
solution of steady-state heat and mass transfer equations. Equations (1) to (7) were used by replacing Kv,ns by Kv,s.
Nevertheless, due to bed stirring, the mass transfer resist-
ance of the bed to solvent vapor was lower and, conse- The Falling Rate Phase
quently, the value of the vapor bed permeability during
Stirring ‘‘Off ’’ Periods
the stirring ‘‘on’’ periods was higher than the vapor bed
The model presented below is based on the well known
permeability during the stirring ‘‘off’’ periods.
phase change front approach first developed by Plank
Consequently, Eqs. (1) to (5) were used by replacing
(Grigull and Sandner[15] and other heat conduction text-
Kv,ns by Kv,s where Kv,s represents the vapor bed per-
books) for 1D semi-infinite medium. In order to extend this
meability during the stirring ‘‘on’’ periods. This parameter
approach to contact drying configuration with intermittent
was obtained by fitting the calculated drying rate curves to
stirring, some modifications of the classical equations were
the corresponding experimental data.
applied. The fundamental basic assumptions were that the
drying rate was entirely controlled by the heat transfer rate
The ‘‘Transition’’ Phase between the heating wall and the bulk of the product and
Stirring ‘‘Off ’’ Periods that a vaporization front moved from the heating wall to
During this phase, the amount of liquid at the heating the bulk of the bed. Between the heating wall and the front,
wall decreased but the solvent vaporization still occurred the product was totally dry and a temperature gradient
only in a thin zone near the heating wall so that the same existed; beyond the vaporization front the temperature
equations corresponding to the constant rate phase were (Tsat,vessel) and solvent content (Xfront) were assumed uni-
still used. Nevertheless, for bed solvent contents ranging form and constant, equal to their values at the beginning
from the first critical solvent content to the second critical of the stirring ‘‘off ’’ period. The main characteristic of vac-
value, noted Xcr2, the heat transfer coefficient between the uum contact drying with intermittent stirring was a strong
 EXPERIMENTAL STUDY OF CRYSTALLINE POWDERS 1167

The thermal energy balance at the vaporization front

was expressed by the following steady-state relationship:
q_ front
_ ¼
m ð11Þ
DHvap
By combining Eqs. (10) and (11), the solvent vapor flux
could be calculated by the following expression:
ðTwall  Tsat;vessel Þ
_ ¼
m ð12Þ
DHvap  ðh1ws þ zfront
kd Þ

which also represented the drying rate.

As well, the solvent balance at the vaporization front
was expressed by the following relationship:
dzfront
FIG. 2. Scheme of temperature and solvent content profiles through the _ ¼ qd  Xfront 
m ð13Þ
dt
Downloaded by [University of Connecticut] at 07:20 12 October 2014

bed during the stirring ‘‘off’’ periods and during the fictitious rest steps of
stirring ‘‘on’’ periods. Moreover, the evolution of solvent content with time was
still described by Eq. (5).
increase of the drying rate at the beginning of each stirring By combining Eqs. (12) and (13), the vaporization front
‘‘on’’ period as observed experimentally. This experimental velocity could be calculated by the following equation:
phenomenon was probably due to the increase of the wall tem- 0 1
perature during the stirring ‘‘off ’’ periods. The thermal inertia dzfront @ 1 A
¼
of the wall resulted in a high temperature gradient in the pro- dt qd  Xfront  DHvap  ð 1 þ Zfront Þ
hws kd
ð14Þ
duct during a short time after the beginning of the stirring
‘‘on’’ periods. Thus, it appeared necessary to take into account  ðTwall  Tsat;vessel Þ
the thermal inertia of the heating wall. The scheme of tem- The parameter value of mwallCpwall was obtained by fitting
perature and solvent content profiles through the bed during the simulated curves to the experimental ones.
the stirring ‘‘off ’’ periods is presented in Fig. 2.
The thermal energy balance at the heating wall was
expressed by the following relationship: Stirring ‘‘On’’ Periods
dTwall The model presented below is based on the so-called
mwall Cpwall ¼ Aðq_ f  q_ ws Þ ð8Þ penetration model comprehensively described in the papers
dt
of Schlünder and Mollekopf [1] and Tsotsas and Schlünder.[2]
where mwall represents the mass of the heating wall, Cpwall The basic principle of the ‘‘penetration’’ model was to
the mass thermal capacity of the heating wall, Twall the apply the convenient ‘‘vaporization front’’ and ‘‘‘series
temperature of the heating wall, q_ f the heat flux coming resistances’’ approach developed for stirring ‘‘off’’ periods
in the heating wall, and q_ ws the heat flux coming out the to a stirred bed drying by substituting the real continuous
heating wall. stirring process by a fictitious sequence of rest steps during
The heat flux coming in the heating wall is expressed by which the bed was static and then interrupted by instan-
the following relationship: taneous adiabatic and perfect mixing of the bed. The
q_ f ¼ hf ðTf  Twall Þ ð9Þ scheme of temperature and solvent content profiles
through the bed during the fictitious rest steps of the
By neglecting the heat accumulation in the dry layer, the heat stirring ‘‘on’’ periods is also presented in Fig. 2.
coming out the heating wall was equal to the heat flux at the The fundamental parameter of this model was the
vaporization front ðq_ front Þ and could be expressed by means fictitious rest step duration (tR), which accounts in a
of the contact thermal resistance at the heating wall and of lumped way for particles trajectories induced by the stirrer
the conduction thermal resistance of the dry layer: motion and which had to be estimated for each particular
0 1 dryer=powder system. In the case of a stirred bed, the
1 thermal contact resistance at the heating wall remained
q_ ws ¼ q_ front ¼ @ A  ðTwall  Tsat;vessel Þ ð10Þ
1 þ Zfront constant, but the thermal conduction resistance within
hws kd
the bed varied according to the time variation of the vol-
where Tsat,vessel represents the saturation temperature at the ume fraction and the spatial distribution of dry and wet
total operating pressure. particles. Thus, this conduction thermal resistance should
 1168 MICHAUD ET AL.

depend on the real trajectory and on the mixing process of case when the drying was operated after a fil-
the particles inside the vessel. tration step, then only two parameters (Nmix and
During the fictitious rest step of a stirring ‘‘on’’ period, the mwallCpwall) had to be fitted instead of the six fit-
basic hypotheses were the same as the ones used for stirring ting parameters (Kv,ns, Kv,s, Xcr1, Xcr2, Nmix, and
‘‘off ’’ periods. However, the drying rate during a fictitious rest mwallCpwall) for the general case;
step corresponded to the average drying rate of the step and
although the dryer heating wall consisted of a plate
the solvent content at the vaporization front (Xfront) was equal bottom part and a cylindrical lateral part, a mono-
to its value at the beginning of the fictitious rest step. The directional heat and mass transfer between the
Eqs. (8), (11), and (13) were used to determine the instan- heating wall and the bulk of the bed was supposed.
taneous values of Twall, ṁ, and zfront. Then, by numerical inte- As concerns heat transfer, the powder was con-
gration of ṁ values over the total duration of the fictitious rest sidered to be uniformly distributed on a plane sur-
step (tR), the average solvent vapor flux or the average drying face whose area corresponded to the surface really
rate ðm _ Þ could be calculated by the following expression: in contact with the powder. As concerns mass trans-
Z fer, the depth of the bed (zmax) corresponded to the
1 tR
_ ¼
m m_ dt ð15Þ experimentally depth of the bed inside the dryer;
tR 0
to simulate the drying rate curves during vacuum
Downloaded by [University of Connecticut] at 07:20 12 October 2014

In the frame of the ‘‘penetration’’ drying model, the transition contact drying with continuous stirring, the model
from one fictitious rest step (k) to the next one (k þ 1) was previously presented for intermittent stirring was
assumed to take place by an instantaneous perfect mixing used by considering only stirring ‘‘on’’ periods and
resulting in uniform solvent content over the entire bed. Conse- by neglecting the thermal inertia of the heating wall.
quently, the bulk solvent content at the beginning of the next
(k þ 1) rest step may be calculated by the following relation:[1] RESULTS AND DISCUSSION

tR A Identifications of Fitted Parameters

Xfrontðkþ1Þ ¼ Xðkþ1Þ ¼ XfrontðkÞ  _k
m ð16Þ Continuous Stirring
md
The experimental and simulated curves corresponding
Moreover, the fictitious rest step duration was equal to the to operating conditions of run 1 are presented in Fig. 3.
ratio of a fitting parameter called the mixing number (Nmix) By fitting the four parameters values of our model (fit-
and the stirrer rotation rate (Z), according to the following ted values given in Fig. 3), a good agreement between
equation:[1] experimental and simulated curves was obtained over the
Nmix total duration of the drying process, including the constant
tR ¼ ð17Þ rate, the transition, and falling rate phases.
Z
The mixing number depended only on the geometrical shape of
Intermittent Stirring
the stirrer, on the vessel geometry, and also on the rheological
The experimental and simulated curves corresponding
(flow) properties of the powder. It represented the number of
to operating conditions of run 2 are presented in Fig. 4.
stirrer revolutions needed to obtain a perfectly mixed bed.
By fitting the six parameters values of our model (fitted
An empirical correlation for this mixing number for some basic
values given in Fig. 4) a good agreement between experi-
contact dryers was proposed by Schlünder and Mollekopf,[1]
mental and simulated curves was observed over the total
but it was more precisely determined for a given dryer=powder
duration of the process, including all three drying phases.
powder system by identification from experimental drying
As well, the identified (fitted) values of the solvent vapor
curves.
(ethanol) bed permeabilities, the critical solvent contents,
The drying rate curve ½m _ ¼m _ ðXÞ and the solvent con-
the mixing number, the mass and thermal capacity of the
tent curve [X ¼ X(t)] over all the stirring ‘‘on’’ period could
heating wall for each trial are compiled in Table 4.
be calculated by combining Eqs. (8)–(10), (12), and (14)–
Nevertheless, a significant difference between the first
(17) from one fictitious rest step to another.
critical solvent content obtained for continuous stirring
The above given model description was completed by
and for intermittent stirring was observed. This difference
the following remarks concerning all the three drying
could probably result from different structures of the
phases:
powder bed in the two stirring cases.

only dry product thermophysical properties, more
easy to obtain than the wet product properties, Model Validation
were involved in the equations model; Our vacuum contact drying model was validated by con-

if the drying period started at an initial solvent ducting three intermittent stirring experiments (trials 3, 4,
content lower than Xcr2, which was generally the and 5) and the corresponding simulations with operating
 EXPERIMENTAL STUDY OF CRYSTALLINE POWDERS 1169
Downloaded by [University of Connecticut] at 07:20 12 October 2014

FIG. 3. Measured and simulated drying rate curves of trial 1 (continuous stirring).

FIG. 4. Measured and simulated drying rate curves of trial 2 (intermittent stirring).

TABLE 4
Identified parameters values of trial 1 (continuous stirring) and trial 2 (intermittent stirring)
Identified parameters values Trial 1 Trial 2
Vapor bed permeability during stirring ‘‘on’’ periods: Kv,s (m2) 0.8  10 11 0.8  10  11
Vapor bed permeability during stirring ‘‘off’’ periods: Kv,ns (m2) — 0.8  10  12
First critical solvent content: Xcr1 0.105 0.125
Second critical solvent content: Xcr2 0.075 0.075
Mixing number: Nmix 15.95 15.95 (from Trial 1)
Mass and thermal capacity of the heating wall: mwallCpwall (J kg 1) — 2400
 1170 MICHAUD ET AL.
Downloaded by [University of Connecticut] at 07:20 12 October 2014

FIG. 5. Measured and simulated drying rate curves of trial 3 (intermittent stirring).

conditions different from the ones used for model entire duration of the process, including all the drying
parameters identification (trials 1 and 2). phases, which represent a coherent validation of our model
The simulated drying rate curves corresponding to trials of vacuum contact drying with intermittent stirring.
3, 4, and 5 presented respectively in Figs. 5–7 were obtained Moreover, the data of Table 5 showed that the overall
by using the same values of the fitting parameters (Kv,s, drying times needed to achieve a fixed final solvent content
Kv,ns, Xcr1, Xcr2, mwallCpwall, and Nmix) than the ones ident- of 0.5% (starting from a solvent content of 25.0%) were
ified with the curves of trials 1 and 2. also very well predicted by our modeling, which repre-
According to Figs. 5–7, a satisfactory agreement between sented a new validation of the precision and adequacy of
experimental and simulated curves was obtained over the our modeling.

FIG. 6. Measured and simulated drying rate curves of trial 4 (intermittent stirring).
 EXPERIMENTAL STUDY OF CRYSTALLINE POWDERS 1171
Downloaded by [University of Connecticut] at 07:20 12 October 2014

FIG. 7. Measured and simulated drying rate curves of trial 5 (intermittent stirring).

TABLE 5 rate data have been obtained by using the same values of
Predicted and experimental drying times for final solvent the model fitting parameters (mwallCpwall and Nmix) as the
content equal to 0.5% from a solvent content of 25.0% ones identified with the experimental data for trials 1 and 2.
(trials 3, 4 and 5) Firstly, the final powder solvent contents after 60 min of
drying for different durations of alternate stirring ‘‘on’’ and
Predicted drying Experimental drying stirring ‘‘off ’’ periods were calculated and the values were
Trials times (min) times (min) reported in the Table 7. The cumulated stirring time was
3 45.7 42.9 always equal to 33% of the total drying time for all
4 69.5 71.2 simulated cases. Consequently, the level of attrition should
5 58.7 59.1 be about the same for all the simulations.
As expected, it can be concluded from these simulated
data that the shorter the stirring on=off cycle was, the
Drying Simulations with Intermittent Stirring lower the final solvent content was and, consequently, the
More potassium chloride drying simulations with inter- higher the overall drying rate was.
mittent stirring were carried out in order to investigate the Moreover, the final powder solvent contents after
impact of the stirring mode on the overall drying rate. The 45 min of vacuum contact drying for a single stirring
operating conditions used for these new simulations except ‘‘off ’’ period of 15 min operated at different drying times
the stirring mode are reported in Table 6. were also calculated and these values were gathered in
The initial solvent content was chosen equal to Xcr2 for Table 8.
that the simulations conditions were close that the operating
conditions of industrial process where the drying is carried TABLE 7
out after the filtration step. The simulated overall drying Calculated solvent content after 60 min of potassium
chloride drying (Xini ¼ 7.5%) for different sequences
of intermittent stirring
TABLE 6 Solvent content
Operating conditions of simulations achieved after
Operating conditions Simulations Simulations 60 min of drying

Mean temperature of hot fluid: Tf ( C) 30 Stirring on=off time: 1.25 min=2.5 min 0.1
Operating total pressure: Pvessel (Pa) 7000 Stirring on=off time: 2.5 min=5 min 0.3
Initial solvent content: Xini 7.5% Stirring on=off time: 5 min=10 min 0.5
Stirring rate: Z (rpm) 23 Stirring on=off time: 10 min=20 min 0.8
 1172 MICHAUD ET AL.

TABLE 8 corresponds to the best compromise between the highest

Calculated solvent content after 45 min of potassium value of the overall drying rate and the least attrition
chloride drying (Xini ¼ 7.5%) for a stirring ‘‘off’’ period phenomena, which will be studied in the last part of our
of 15 min operated at different drying times project.
Solvent content
achieved after NOMENCLATURE
Simulations 45 min of drying (%) A Heating wall area in contact with the product (m2)
Stirring ‘‘off’’ period at t ¼ 0 min 0.9 Cp Mass thermal capacity (J kg 1 K 1)
Stirring ‘‘off’’ period at t ¼ 15 min 1.1 dpa Particle mean diameter (m)
Stirring ‘‘off’’ period at t ¼ 30 min 1.4 4Hvap Latent heat of vaporization (J kg 1)
hf Overall heat transfer coefficient of the jacketed
vessel (W m 2 K 1)
Thus, it could be concluded from these simulation data hg Overall heat transfer coefficient (W m 2 K 1)
that the stirring ‘‘off ’’ period should be placed at the begin- hws Contact heat transfer coefficient (W m 2 K 1)
ning of the drying process in order to achieve the highest K Solvent bed permeability (m2)
ṁ Vapor flux or drying rate (kg m 2s 1)
Downloaded by [University of Connecticut] at 07:20 12 October 2014

value for the overall drying rate.

m_ Average vapor flux or average drying rate
(kg m 2 s 1)
CONCLUSIONS
m Mass (kg)
Most of literature data concerning modeling of vacuum Nmix Mixing number
contact drying with continuous stirring are based on the P Pressure (Pa)
‘‘penetration’’ model firstly proposed by Schlünder and
Pvessel Operating total pressure (Pa)
Mollekopf.[1] This classical model was also the basic
q̇f Heat flux coming in the heating wall (W m 2)
approach of the present work aimed at the modeling and
q̇front Heat flux at the vaporization front (W m 2)
simulation of the drying rates and drying times in the case
q̇ws Heat flux coming out the heating wall (W m 2)
of vacuum contact drying with intermittent stirring. T Temperature (K)
Compared with the literature, our modeling approach t Time (s) or (min)
incorporated the following modification: tR Rest step duration (s)

the introduction of a jacketed vessel heat transfer X Averagesolvent content (dry basis, kg solvent
coefficient and an accumulation term for heating kg 1 dry product)
wall; Xcr1 First critical mean solvent content (dry basis,

the application of alternate static bed and stirred kg solvent kg 1 dry product)
bed conditions; Xcr2 Second critical mean solvent content (dry basis,

the modeling of all the three drying phases, namely kg solvent kg 1 dry product)
the constant rate phase, the transition phase, Z Stirring rotation rate (s 1)
and the falling rate phase. z Distance from the heating wall (m)
zmax Depth of the bed (m)
Several fitting parameters were used, but only two para-
meters values were necessary to implement our model if the Greek Letters
drying process consisted only of the falling rate phase.
g Absolute viscosity (kg m 1 s 1)
In order to validate our extended model, several experi-
k Thermal conductivity (W m 1 K 1.
ments were carried out with operating drying conditions q Apparent density (kg m 3)
differing from the ones used for identification of parameter w Porosity
values. Drying rates and temperature profiles for a test pro-
duct (potassium chloride) and for different operating con- Subscripts
ditions (pressure, heating fluid temperature, duration of cr Critical
stirring ‘‘on’’ and stirring ‘‘off’’ periods) were experimen- d Dry product or dry layer
tally determined. A good agreement between the simulated f Heating fluid
and experimental curves was observed and the experi- front Vaporization front
mental drying times were also very well predicted. ini Initial
By using our validated physical drying model, other dry- k kth Rest step
ing simulations for different sequences of stirring and no mix Mixing
stirring periods were realized. As well, the final objective ns Stirring ‘‘off ’’ periods (no stirring)
of this study is to find an optimal stirring sequence that s Stirring ‘‘on’’ periods
 EXPERIMENTAL STUDY OF CRYSTALLINE POWDERS 1173

sat Saturation 5. Heimann, F.; Schlünder, E.U. Vacuum contact drying of mechanically
agitated granulate beds wetted with a binary mixture. Chemical
tran Transition period Engineering Processing 1988, 24, 75–91.
v Solvent vapor 6. Farges, D.; Hemati, M.; Laguérie, C.; Vachet, F.; Rousseaux, P. A new
vap Vaporization approach to contact drying modeling. Drying Technology 1995, 13,
vessel Vessel 1317–1329.
wall Heating wall 7. Dittler, A.; Bamberger, T.; Gehrmann, D.; Schlünder, E.U. Measure-
ment and simulation of the vacuum contact drying of pastes in a
LIST-type kneader drier. Chemical Engineering Processing 1997, 36,
301–308.
ACKNOWLEDGEMENTS 8. Tsotsas, E. Über den einfluß der dispersität und der hygroskopizität auf
This work was financially supported by the French den trocknungsverlauf bei der vakuum-kontakttrocknung rieselfähiger
trocknungsgüter; Dissertation, Universität Karlsruhe, Karlsruhe, 1985.
Agency for Environment and Energy Management ADEME
9. Moyne, C. Transferts couples chaleur—masse lors du séchage: prise en
and the pharmaceutical group SANOFI-AVENTIS. compte du mouvement de la phase gazeuse; Thèse, Institut National
Polytechnique de Lorraine, Nancy, 1987.
10. Baillon, B. Séchage sous vide micro-ondes combiné de granulés pharma-
ceutiques; Thèse, Université de Pau et des Pays de l’Adour, Pau 1996.
REFERENCES 11. Kohout, M.; Collier, A.P; Stepanek, F. Mathematical modelling of
Downloaded by [University of Connecticut] at 07:20 12 October 2014

1. Schlünder, E.U.; Mollekopf, N. Vacuum contact drying of free flow- solvent drying from a static particle bed. Chemical Engineering
ing mechanically agitated particulate material. Chemical Engineering Science 2006, 61, 3674–3685.
Processing 1984, 18, 93–111. 12. Kohout, M.; Collier, A.P; Stepanek, F. Multi-scale analysis of
2. Tsotsas, E.; Schlünder, E.U. Vacuum contact drying of free flowing vacuum contact drying. Proceeding of IDS 2006, Vol. B, 1087–1094.
mechanically agitated particulate multigranular packing. Chemical 13. Schlünder, E.U. Wärmeübergang an bewegte kugelshüttungen
Engineering Processing 1986, 20, 339–349. bei kurfristigem kontakt. Chemie Ingenieur Technik 1971, 43,
3. Tsotsas, E.; Schlünder, E.U. Vacuum contact drying of mechanically 651–654.
agitated beds: The influence of hygroscopic behaviour on the drying 14. Mollekopf, N.; Martin, H. Zur Theorie des wärmeübergangs an bewegte
rate curve. Chemical Engineering Processing 1987, 21, 199–208. kugelschüttungen bei kurzfristigem kontakt. VT-Verfahrenstechnik
4. Robert, R.; Heimann, E. Vacuum contact drying of mechanically 1982, 16, 701–706.
agitated coarse, hygroscopic bulk material. Chemical Engineering Pro- 15. Grigull, U.; Sandner, H. Heat Conduction; Hemisphere Publishing
cessing 1989, 26, 225–235. Corporation, USA: 1984.