You are on page 1of 12

Services on Demand

Article
pdf in English
Article in xml format
Article references
How to cite this articleCurriculum ScienTI
Automatic translation
Send this article by email
Indicators
Cited by SciE!" Access statistics
#elated lin$s
Similars in SciE!"
%oo$mar$
Share on deliciousShare on googleShare on twitterShare on diggShare on
citeuli$eShare on connotea&'ore Sharing Services'ore

(ermalin$
%ra)ilian *ournal of Chemical Engineering
(rint version ISS+ ,-,.//01
%ra)2 *2 Chem2 Eng2 vol211 no21 S3o (aulo Apr24*une 1,,5
http644dx2doi2org4-,2-57,4S,-,.//011,,5,,,1,,,-8
D#9I+: SCIE+CE A+D TECH+"!":9

'odeling and simulation of mil$ emulsion drying in spray dryers


;2 S2 %irchalI< '2 !2 (assosI= II= >


I(rograma de (?s:radua@3o em Engenharia 'ecAnica= Bniversidade Cederal de
'inas :erais= Av2 AntDnio Carlos //1E= 0-1E,7,-= %elo Hori)onte ':= %ra)il E
mail6 vbirchalFmatrix2com2br
IICentro de Secagem= Departamento de Engenharia GuHmica= Bniversidade Cederal de
S3o Carlos= (2"2 %ox /E/= -05/57,5= S3o Carlos S(= %ra)il Email6
merilauFmicroplanet2com2br

A%ST#ACT
This wor$ aims at modeling and simulating the drying of whole mil$ emulsion in
spray dryers2 Drops and particles ma$e up the discrete phase and are distributed into
temporal compartments following their residence time in the dryer2 Air is the
continuous and wellmixed phase2 'ass and energy balances are developed for each
phase= ta$ing into account their interactions2 Constitutive eIuations for describing the
drop swelling and drying mechanisms as well as the heat and mass transfer between
particles and hot air are proposed and analy)ed2 A set of algebraicdifferential
eIuations is obtained and solved by specific numerical codes2 #esults from
experiments carried out in a pilot spray dryer are used to validate the model developed
and the numerical algorithm2 Comparing the simulated and experimental data= it is
shown that the model predicts well the individual dropparticle history inside the
dryer as well as the overall outlet airparticle temperature and humidity2
Jeywords6 Spray dryer operation< (article distribution< Drying model< +umerical
simulation2

I+T#"DBCTI"+
The industrial process for producing powdered mil$ comprises the basic steps shown
in Cigure -2 #eceipt of raw mil$ from farms or cooperatives involves its inspection
based on legal regulation of chemical= sensorial and bacteriological compositions2
After approval in this initial control= mil$ emulsion is clarified in centrifugal
separators or filters= cooled in heat exchangers and stored in tan$s at .KC2 In the
standardi)ation step= the ratio of mil$ fat to total solids is adLusted= as reIuired for the
final product2 Heat treatment= which includes pasteuri)ation= sterili)ation and
ultrahigh temperature heating= destroys pathogenic microorganisms and produces
physicochemical changes in raw mil$ to improve its shelf life2 The evaporation step=
usually carried out continuously at ., to E,KC in a multieffect evaporator system=
increases the mil$ solids content from -1 M to about 5, M2 The following step
comprises a single= two or multistage drying= in which the concentrated mil$
emulsion is dried and powdered until reaching the residual moisture content reIuired
for storage and4or further application2


Spray drying is a well$nown industrial techniIue used on a large scale for drying and
powdering very thermally sensitive materials2 Cor mil$ emulsion= this techniIue
transforms this emulsion into a large number of small droplets that fall into the spray
chamber concurrently with hot air2 As water is evaporated= these droplets become
solid particles2 In singlestage drying= using a spray dryer with a pneumatic conveying
system= removal of the last fraction of moisture from powdered mil$ occurs slowly
and is costly= and the final product is fine nonagglomerated particles N0,5, OmP that
do not readily disperse in water2 A spray dryer attached to a vibrating fluid bed
system= as shown in Cigure -= comprises the twostage drying2 In this method=
agglomerates are preserved at the exit of the spray chamber and= thus= feed into the
vibrofluidi)ed bed system where they are gently dried and cooled2 +onagglomerated
particles can be recycled into the spray chamber= enhancing the Iuality of powdered
mil$ by control of the particle si)e distribution2 Threestage drying= in which a
stationary fluid bed of powdered mil$ is added at the conical base of the spray
chamber to better control particle agglomeration and drying= is an improvement of
this method2 After drying and cooling= the final powdered mil$ product must be
pac$aged in suitable containers and stored= when necessary= under conditions specific
to its safe preservation2
"f all these steps= drying is the one in which powdered mil$ properties are defined2
As is well$nown in the literature= powdered mil$ and its byproducts have
characteristics crucial to its acceptance and commerciali)ation2 The Iuality of
powdered mil$ is greatly affected by the operational conditions of this process2
Therefore= the $nowledge of the spraydrying technology is essential for controlling
these operational variables in order to obtain a highIuality product NJing et al2= -78.<
+ath and Satpathy= -778P2
%ased on the previous mathematical model presented by Clement et al2 N-77-P= which
focuses on Euler and !agrange approaches for developing mass and energy balances=
this wor$ aims at modeling and simulating the drying of whole mil$ emulsion in spray
dryers2
As shown by the results= a more detailed study of the mass transfer model and the
particle residence time distribution has improved the original Clement et al2 model2 A
new and more efficient method for the numerical model solution was implemented2
Simulations allowed evaluation of model applicability= study of operational parameter
sensitivity and prediction of the dryer outlet conditions at several levels of operational
parameters2

'ATHE'ATICA! '"DE! C"# D#9I+: 'I!J E'B!SI"+ I+ S(#A9
D#9E#S
Initial Assumptions
The mathematical formulation is based on the following assumptions NClement et al2=
-77-P6
Qhot gas and mil$ suspension are continuously inLected into the top of the dryer
chamber at uniform rates<
Qthe gas phase= formed of air and vapor= behaves as an ideal mixture and flows inside
the dryer as a perfect mixture<
Qthe suspension is properly atomi)ed forming spherical= homogeneous and uniform
drops= which are well mixed inside the chamber without interacting with one another<
Qheat flows from hot air to drops or particles and temperature gradients inside the
particles are negligible<
Qthe individual drying rate is described by water evaporation transfer from a single
particle to the gas phase<
Qthe overall drying rate is a weight sum of individual rates of all particles that remain
in the dryer<
Qthe drying $inetics is represented by two distinct periods6 the first one= in which free
water is evaporated from drops until a thin semipermeable crust is formed on the
particle surface= and the second one= characteri)ed by the thic$ening of this crust as
predicted by the shrin$ing core model2
'odel Development
The model comprises a set of constitutive algebraic eIuations that describes mass and
heat transfer between gas and single particles= the drying $inetics of a single particle
surrounded by gas and functions of the particle residence time distribution inside the
chamber< a group of differential eIuations that expresses the mass and energy
balances for a single particle and a continuous gas phase and a set of algebraic
eIuations that express the overall distributed balances of particles inside the chamber2
In this wor$= only the main model eIuations are presented< complete information
about the entire model is given by %irchal N1,,0P2 Constitutive algebraic eIuations
concerned with a single particle can be summari)ed as
Heat and mass transfer rates per unit of the particle surface area R
with
'ean particle diameter R dp
'ass of dry solids inside the drop R ms
EIuilibrium moisture content R S>
where particle temperature= Tp= is expressed in J2
EIuations N-P and N1P respectively describe the heat and the mass transfer between a
particle and gas2 Cor estimation of the convective heat and mass transfer coefficients=
h and $= +usselt N+uP and Sherwood NShP numbers are supposed to be eIual to 1=
since the relative velocity between particle and gas is nearly )ero2 The f factor in
EIuation N1P represents the resistance of the crust to water vapor diffusion through it=
and this value arises only in the second drying period NSp T ScP2 #ight at the
beginning of the first drying period= even when the particle temperature= Tp= becomes
greater than the wetbulb temperature= the drying rate is adLusted by an additional
term to ma$e the vapor pressure in the bul$ air= (wg= eIual to the saturated vapor
pressure at wetbulb temperature= (w>NT%BP2 This additional drying rate per particle
area= = is thus included in the modified model= as shown in EIuation N--P2
EIuations N0aP and N0bP describe the particle4drop diameter during drying2 The initial
drop diameter= dp,= a function of atomi)er design parameters and the suspension flow
rate= is estimated by the specific correlation for the rotary disc atomi)er N'asters=
-785P2
According to the model assumptions= in the second drying period this diameter
becomes constant2 The mass of dry solids inside a particle= ms= constant throughout
the process= can be obtained from EIuation N.P2 EIuation N5P describes the particle
eIuilibrium moisture content= S>= as a function of particle temperature= Tp= and
relative humidity of the air= B#2 This sorption curve= specific for the whole powdered
mil$ used in this wor$= was obtained from experiments carried out under dynamic
conditions N%irchal= 1,,0P2 This curve is also used to specify the critical particle
moisture content= Sc= by ma$ing B# close to -2 +ote that Sc depends on Tp because
of the crust formation on the particle surface2
To determine the mass and energy balances of the particle population in the dryer
Ndiscrete phaseP= it is necessary to specify the effective contribution of each particle or
particle group to the overall drying and heat exchange rates2 This contribution is
directly related to the residence time of the particle in the dryer2 Hence= a detailed and
clear description of the residence time distribution function is extremely important2
The residence time distribution function of particles in the dryer is expressed as a
function of the cumulative fraction of particles which leaves the dryer at time t= CcNtP=
or even as a function of the age distribution of these particles= ENt4trpP= in relation to
the mean particle residence time= trp2 In this wor$= the particle residence time
distribution is based on the perfect mixture model< however= different models for the
particle residence time distribution can be easily incorporated into the model program2
%ased on this assumption= particles fed into the chamber at the initial time Nt U ,P are
distributed along different routes in the dryer and are removed at different times= as
predicted by
Since particles ta$e different routes inside the drying chamber= it is necessary to
distinguish= at any operating time t= that particle fraction of ages less than or eIual to
t= which is removed from the dryer and that particle fraction of ages greater than to t=
which is $ept inside the chamber2 It is also necessary to differentiate= in these
fractions= the correct particle age in order to determine the water evaporation rate and
the rate of heat exchanged with the gas phase2 This reIuires combining sameage
particles into n temporal compartments= which can simplify identification during the
operating time= the remaining and removed fractions of particles according to age2
This approach of temporal compartmentali)ation is Lustified= since in the drying
model= particles of the same age have the same diameter= moisture content and
temperature2 +ote that= in the model assumptions= drop coalescence and drop
brea$age are both neglected2 Therefore= particles of the same age= belonging to the
same temporal compartment= can be found in different spatial positions in the drying=
as represented in Cigure 12


Bsing this approach= the disperse phase is thus supposed to be comprised of particles
that are distributed into n temporal compartments by age2 Bnder inlet condition t U ti
V , Ni U 1 to nP= the particle fraction that remains in the chamber at a time t V ti is
expressed by
EIuation NEP allows calculation of the fraction of particles of an age between ti and Nti
W dtP= which remain in the dryer at operating time t2 Cigure 0 illustrates the model
compartments corresponding to the particle age at any time t U i Dt N, T i T nP2 +ote
that at t U Nn-PDt= the n compartments are all filled with particles of ages varying
from , to tn2 'oreover= at this operating time= the remaining fraction of particles fed
in at t U , leaves the dryer2 If there is no perturbation in this dryer system= the
permanent regime is established at t U nDt2 The +p=-=L NU DtC4msP compartment is
composed of particles of age t- U , at t U NL -PDt for - U L U n< +p=1=L NU r- +p=-=L-P
by particles of age t1 U Dt at t U NL -PDt for 1 U L T n< +p=0=L NU r1 +p=1=L-P by those
of age t0 U 1Dt at t U NL -PDt for 0 U L U n and so on consecutively until +p=n=L NU rn-
+p=n-=L-P is reached at t U Nn-PDt2 Cor any i value= ri U exp NDt4trpP is constant2
Crom this matrix scheme in Cigure 0= compartments composed of particles fed into the
dryer at t U , are on the principal diagonal Ni=L U iP corresponding to +p=i=LUi2
Compartments with particles fed into the dryer at t U Dt are located on the diagonal Ni=
L U i W -P= corresponding to +p=i=LUiW- and so on successively2 In each compartment=
there are two outlet currents of particles6 one formed by the particle fraction +p=s U
N-riP+p=i=L that leaves the dryer at NL -PDt U t U L Dt and the other by the particle
fraction +p=r U ri+p=i=L that remains in the dryer and feeds the next time compartment=
+p= iW-=LW-2


'oreover= this scheme= different from the one suggested by Clement et al2 N-77-P=
allows easy identification of the particle fraction that leaves the dryer at t V Dt2 %ased
on this= water evaporation and the heat transfer rates can be determined during the
entire operating time= since the individual mass and heat transfer rates estimated in
each i of the +pi=L compartments in the L column of a given operating time= can be
summed up as follows6
"verall water evaporation rate
"verall heat rate
+p=i=L is the particle fraction in the i=L compartment N- U i U nP with L U constant
corresponding to a given operating time and expressed as
with ti U Ni R -PDt2
%ased on EIuations N8P and N7P= mass and energy balances for both solid and gas
phases were developed as follows6
'ass balance for a single particle Napplied to n compartmentsP
Energy balance for a single particle Napplied to n compartmentsP
'ass balance for the gas phase
Energy balance for the gas phase
This set of balances= from EI2 N--P to N-.P= comprises N1nW1P differential eIuations2
"verall rates of mass and heat transfer between phases= the mean particle moisture
content= and temperature== can thus be calculated at a given operating time= t= as
"verall heat rate supplied by gas
"verall heat rate used for water evaporation
"verall mass transfer rate for the solid phase
'ean particle moisture content in the dryer
'ean particle temperature in the dryer
This modified model has two adLustable parameters= which are Diff= the effective
coefficient of watervapor diffusion through the particle crust= and hwall= the overall
coefficient of heat transfer between dryer walls and the ambient air2 These
coefficients are estimated from experimental data= as discussed later2

+B'E#ICA! 'ETH"D C"# S"!;I+: '"DE! EGBATI"+S
Due to the complexity of the differentialalgebraic eIuation system NDAEP= its
solution reIuires the use of numerical methods2 Transforming this system into one of
ordinary differential eIuations N"DEP would involve a series of algebraic operations=
such as the one used by Clement et al2 N-77-P2 %esides= these authors reported serious
convergence problems while simulating the beginning of the drying process2 These
arguments are sufficient to support the proposal of a new method to solve directly this
DAE system= in which the DASS! code N(et)old= -787P is applied2 %y using the
bac$warddifferential formula in each integration step= the DASS! code transforms
DAE system into a set of algebraic eIuations= whose solution is obtained by +ewton
methods2 The code= developed in the C"#T#A+ language= has been incorporated into
a computational C"#T#A+ program= containing the initial conditions< the seIuential
order of eIuations to be solved with the corresponding code callings and the solution
feedbac$ to the program2 In addition= there are specific subroutines for calculating the
air= vapor= water and drop properties as well as the eIuilibrium data on both phases
lin$ed to the main program2
In this model solution= the convergence problem was solved by using different
discreti)ation times during the simulation2 Since= in the first temporal compartment
N-=-P= values and tendencies of temperature and gas humidity are un$nown= problems
with the DASS! initiali)ation can emerge2 %esides= at the beginning of the drying
process= there are fast and abrupt changes in the particle temperature and moisture
content profiles2 Therefore= it is necessary to implement a large number of
compartments Nn V -,,,P in the first drying period2 However= once convergence is
achieved= this large number becomes unnecessary2 To optimi)e the model solution
algorithm= a time refining in the first temporal compartment is proposed= subdividing
it into nr temporal subcompartments with a time interval of Dt4nr2 Simulations have
shown that setting the number of compartments at n U 5,, and nr U 5, is the
minimum reIuired to guarantee convergence2 In this wor$= n U /,, and nr U 1,, were
adopted2 The results= as shown in Cigure .= are a good description of particle history
inside the dryer2


To estimate the two adLustable model parameters= Diff and hwall= the weighty least
sIuare method was used to represent the residue or error function2 An optimi)ation
subroutine= based on the (S" Nparticle swarm optimi)ationP heuristic method
NJennedy and Eberhart= -775P was implemented= resulting in a rapid convergence to
the overall minimum of this error function2 As a result= Diff was estimated as .2-x-,
7 m14s and hwall as E20 S4m1J2

#ESB!TS A+D DISCBSSI"+
To analy)e and validate results from the model developed= experimental tests were
conducted at IE+4C+E+ in an A(;A+HID#" model (SD 51 pilot spray dryer
made of stainless steel with a - m of diameter and a ,2/ m of height Ncylindrical partP
and a E25 $g4h of evaporation capacity2 Drying air at a rate of ,2,.- $g4s and a
humidity of ,2,-, to ,2,1, and whole mil$ emulsion with a .7M solids concentration
at 5-XC are fed in concurrently= reproducing the industrial conditions of dryer
operation2 The atomi)er is of the rotary disc type2 +ine inlet conditions were tested by
combining low N-P= medium N,P and high NW-P levels of three operational variables6 C=
vatom and Tg,= as shown in Table -2 'ore details of these experiments can found in
the wor$ of %irchal N1,,0P2


#esults obtained with the model by using the computer program developed are
presented in Table 1 together with experimental data= deviation and fitted error
Nmodel fit to experimental dataP2 In figures 5a= 5b and 5c= the agreement between
experimental and simulated data can easily be observed2 In these figures= vertical bars
on the experimental data points represent the experimental error given in Table 12


As shown in Cigure 5a= all values of Tg are predicted well by the model within the
experimental uncertainty range2 This variable is one of the most important in the spray
drying operation= since it is an indirect measurement of solids properties= such as the
outlet powder moisture content and temperature in the industrial production of
powdered mil$2 Crom this= it is also possible to calculate thermal dryer efficiency and
to analy)e particle temperature history and possible alterations in properties of the
powdered mil$2
Cigure 5b indicates a larger deviation between experimental and simulated data=
which is explained by the considerable measurement errors obtained with instruments2
The model predicts satisfactorily nine out of eleven data2 !arger and no predictive Sp
deviations obtained in experiments 0 and E may be related to the severe inlet
conditions imposed on the mil$ drying Nhigher temperature and lower emulsion feed
rateP= which resulted in a burned product with a possible loss of some volatile
components2
In Cigure 5c= 9 is also predicted well= considering the experimental uncertainty
involved in measurement of this variable2

C"+C!BSI"+S
%ased on the results obtained in this study= the following conclusions are drawn6
Qthe modified model satisfactorily describes its response variables= Tg= Sp and 9=
under continuous operation<
Qalthough the particle residence time distribution can be improved by testing other
functions= the perfect mixture approach seems to describe reasonably well particle
motion inside the pilot spray dryer used in the experiments<
QTg= one of the most important process variables= has the lowest model predicted
error<
Q9 is also predicted well< nevertheless= the associated experimental uncertainty is high
due to the time reIuired for getting trusty online results<
QSp is satisfactorily predicted by the model in the usual range of inlet air
temperatures NT -8, XCP< however the model cannot predict the thermal degradation of
powdered mil$= as observed in the experiments= at a high level of inlet air
temperature2
Studies are in progress to improve this modified model by designing control strategies
for safely drying mil$ emulsion in spray dryers2

+"'E+C!ATB#E

#ECE#E+CES
%irchal= ;2S2= 'odelagem e simula@3o da secagem de leite em secadores spray2 (h2D2
diss2= Bniversidade Cederal de 'inas :erais= %elo Hori)onte N1,,0P2 Y !in$s Z
Clement= J2H2= Hallstr[m= A2= Dich= H2C2= !e= C2'2= 'ortensen= *2 and Thomsen=
H2A2= "n the Dynamic %ehavior of Spray Dryers2 Transactions Inst2 of Chemical
Engineers= /7= 1.5 N-77-P2 Y !in$s Z
Jennedy= *2 and Eberhart= #2C2= (article Swarm "ptimi)ation= International
Conference on +eural +etwor$s2 (erth= Australia N-775P2 Y !in$s Z
Jing= C2*2= Jiec$busch= T2:2 and :reenwald= C2:2= CoodGuality Cactors in Spray
Drying= Advances in Drying= 0= E- N-78.P2 Y !in$s Z
'asters= J2= Spray Drying Handboo$2 !ongman Scientific and Technical= +ew 9or$
N-785P2 Y !in$s Z
+ath= S2 and Satpathy= #2= A Systematic Approach for Investigation of Spray Drying
(rocess= Drying Technology= -/= +o2 / --E0 N-778P2 Y !in$s Z
(et)old= !2#2= DASS!6 a Differentialalgebraic System Solver= Computing and
'athematics2 #esearch Division= !awrence !ivermore +ational !aboratory=
!ivermore= California N-787P2 Y !in$s Z


#eceived< "ctober 1,= 1,,.
Accepted6 'arch 0= 1,,5


> To whom correspondence should be addressed
All the contents of this Lournal= except where otherwise noted= is licensed under a
Creative Commons Attribution !icense
Associa@3o %rasileira de Engenharia GuHmica
#ua !Hbero %adar?= -51 = --2 and2
,-,,87,0 S3o (aulo S( %ra)il
Tel26 W55 -- 0-,E8E.E
Cax26 W55 -- 0-,../.7
Cax6 W55 -- 0-,../.7
rgiudiciFusp2br