You are on page 1of 11

Journal of Food Engineering 71 (2005) 55–65

www.elsevier.com/locate/jfoodeng

Optimization of co-current spray drying process of sugar-rich


foods. Part I—Moisture and glass transition
temperature profile during drying
a,*
Vinh Truong , Bhesh R. Bhandari b, Tony Howes c

a
Faculty of Food Science and Technology, The University of Agriculture and Forestry, HCM city, Vietnam
b
School of Land and Food Sciences, The University of Queensland, St. Lucia QLD 4072, Australia
c
School of Engineering, The University of Queensland, St. Lucia QLD 4072, Australia

Received 2 February 2004; accepted 7 October 2004


Available online 2 December 2004

Abstract

A steady state mathematical model for co-current spray drying was developed for sugar-rich foods with the application of the
glass transition temperature concept. Maltodextrin–sucrose solution was used as a sugar-rich food model. The model included mass,
heat and momentum balances for a single droplet drying as well as temperature and humidity profile of the drying medium. A log–
normal volume distribution of the droplets was generated at the exit of the rotary atomizer. This generation created a certain num-
ber of bins to form a system of non-linear first-order differential equations as a function of the axial distance of the drying chamber.
The model was used to calculate the changes of droplet diameter, density, temperature, moisture content and velocity in association
with the change of air properties along the axial distance. The difference between the outlet air temperature and the glass transition
temperature of the final products (DT) was considered as an indicator of stickiness of the particles in spray drying process. The cal-
culated and experimental DT values were close, indicating successful validation of the model.
 2004 Elsevier Ltd. All rights reserved.

Keywords: Glass transition temperature; Mathematical modeling; Optimization; Maltodextrin; Sucrose; Spray drying; Stickiness; Sugar-rich foods

1. Introduction stickiness and the design of a dryer system. Theoreti-


cally, there should no deposition problem if the size of
Spray drying of sugar-rich foods is difficult due to the the dryer chamber is large enough. Because of the eco-
presence of high content of sugars and organic acids. nomic factors, the size of the dryer needs to be limited
These compounds exhibit sticky behavior during spray to a suitable range. Therefore, stickiness and deposition
drying. Stickiness is a term used to describe the phenom- need to be solved by other methods rather than by using
ena of particle–particle cohesion and particle–wall adhe- a very large dryer chamber.
sion in the spray drying process. Stickiness depends not Additives such as glucose syrups or maltodextrin
only on the properties of materials but also on the inlet have been used to produce physical changes in the prod-
variables applied in a spray drying system. Deposition is uct, consequently reducing the stickiness and wall depo-
another phenomenon in spray drying. It relates to the sition in spray drying. Glucose syrups have been used
for blackcurrant (Zaleski, Lipowska, & Kuszlik, 1968)
and orange juice (Brennan, Herrera, & Jowitt, 1971)
*
Corresponding author. Tel.: +84 08 8975614; fax: +84 08 8960713. and maltodextrin has been used for different sugar-rich
E-mail address: vinhthao@hcmc.netnam.vn (V. Truong). foods such as orange (Gupta, 1975), blackcurrant

0260-8774/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfoodeng.2004.10.017
56 V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65

Nomenclature

A, B constants in latent heat equation RMS ratio between solid fractions of maltodextrin
a 1, a 2 slopes in falling period of drying rate curve and sucrose
(FPDR) Sc Schmidt number
Ap droplet (particle) surface area, m2 Sh Sherwood number
c 1, c 2 constants in modified Henderson equation T temperature, K or C
Cd drag force coefficient Tg glass transition temperature, K or C
Cp specific heat, J/kg K Tsticky sticky temperature, K or C
d diameter, m (lm) UR relative velocity between particles and air,
da diameter of atomizer disc, m m/s
DgNw geometric mean of number distribution of Ux, Ur, Ut absolute velocity of air/particle in direc-
droplet, lm tions x, r and t, m/s
DgVd, DgVw geometric mean of volume distribution of x mass fraction
dried-particle, droplet, lm X moisture content, w/w dry basis
Dvsd, Dvsw sauter-mean diameter of dried-particle, X0 moisture content at intersection between two
droplet, lm straight lines of falling period of the drying
Dv diffusivity of water in air, m2/s curve
F(j), F(lndp) frequency of number distribution of log– Xc critical moisture content, w/w dry basis
normal distribution Y humidity of air, kg/kg
f(lndp) density function of log–normal distribution DT temperature difference between surrounding
Fv(j) frequency of volume distribution (air, wall) and Tg, C
g gravity acceleration, m/s2 DTpg temperature difference between particle tem-
G mass flow rate of dry air, kg/s perature and Tg, C
h axial distance, m
H enthalpy, J/kg Greek symbols
Hv enthalpy of vapor in air, J/kg n relative drying rate
hfg latent heat of evaporation of pure water, J/kg q density, kg/m3
Hfg latent heat of water evaporation in food, J/kg D difference
hfgdewp latent heat of vaporization at dew point tem- l viscosity, Pa s
perature, J/kg rgd standard deviation of the logarithm of the
Hh enthalpy of humid air, J/kg diameter of dried-particle
hv height of vane of atomizer disc, m rgw standard deviation of the logarithm of the
K constant in Gordon–Taylor equation diameter of droplet
Ka, Kd thermal conductivity of air, particle, W/mK
Km mass transfer coefficient, kg/m2 s Subscripts
Ksh shrinkage ratio a air
m mass of particle/sample, kg b bulk air
Mp mass flow rate per total periphery of rotary c critical
atomizer, kg/ms d dry
Na speed of atomizer, RPM dewp dew point
Ndrop number of droplets generated from atomizer drop droplet
Nu Nusselt number e, L equilibrium, liquid
nv number of vanes in disc of atomizer F feed
Pr Prandtl number film film layer surrounding particle
Ps saturation vapor pressure, Pa g glass transition
Pv partial vapor pressure at film temperature, Pa h humid air
Pv1 partial vapor pressure at bulk air tempera- in inlet
ture, Pa MD, M maltodextrin
QF feed rate, kg/s N number distribution
r radial distance, m out outlet
Re Reynolds number ov overall
RH relative humidity, decimal p particle, product
V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65 57

r radial axis V volume distribution


s solid W water
S, suc sucrose wall wall
T at temperature T wb wet bulb
t tangential axis x axial axis
v vapor

(Bhandari, Senoussi, Dumoulin, & Lebert, 1993), honey of this work. Before optimization of the spray drying
(Bhandari, Datta, Crooks, Howes, & Rigby, 1997) and process can be made as presented in Truong, Bhandari,
tamarind (Truong, 1994). The addition of theses carriers and Howes (2004), a mathematical model is developed
into the feed improved the powder recovery. The frac- in this paper. This model was used to determine mois-
tion of additives in juices is normally in the range of ture and glass transition temperature profile of the drop-
40–60% but higher fraction can also be used (Masters, let/particles during spray drying process.
1994). However, the required sensory quality of final There are several mathematical models in the litera-
products limits the amount of additives. tures proposed for the drying calculation in a co-current
Other methods try to control the surrounding prod- spray dryer. For instance, some spray drying models
uct or air temperatures, such as using appropriate outlet have been applied in the literatures were summarized
drying air temperature (Bhandari, Datta, Crooks, et al., by Crowe (1980). In this classification, models were cat-
1997) or introducing of cold air into the bottom part of egorized as one-dimensional, quasi-one-dimensional or
the dryer (Lazar, Brown, Smith, Wong, & Lindquist, axisymmetric flow. The axisymmetric model is general
1956) or cooling of the wall temperature (Brennan but costly to implement because of its detail. Negiz, Lag-
et al., 1971). The improvement of the spray drying per- ergren, and Cinar (1995) classified dynamic and steady
formance while applying these methods to reduce stick- state models. The dynamic state models (Clement
iness has been reviewed elsewhere (Bhandari, Datta, & et al., 1991; Crowe, Sharma, & Stock, 1977; Langrish &
Howes, 1997). Zbicinski, 1994; Oakley & Bahu, 1993; Papadakis &
Fast evaporation in spray drying produces the parti- King, 1988a, 1988b; Zbicinski, 1995) are concerned with
cles in amorphous form which exhibits the glass transi- the transient behavior of the drying air and droplets.
tion. The high content of low molecular weight sugars The steady state models (Dickinson & Marshall, 1968;
(sucrose, glucose and fructose) and organic acids (tar- Negiz et al., 1995; Parti & Palancz, 1974) deal with
taric, citric and malic acids) depresses the glass transi- steady operation conditions.
tion temperature (Tg) of the material below the Dynamic models are considered mainly for process
product temperature (Tp) even at the exit of the dryer. control. However, since a stable operating condition is
This leads to the existence of the liquid-like state of always desired (for the process to operate), models for
the amorphous material, which is responsible for inter- the steady state behavior of the process also have re-
particle cohesion or particle adhesion on the dryer sur- ceived attention (Negiz et al., 1995). The steady state
faces. The higher the temperature difference one-dimensional model was initially developed by Parti
(DTag = Ta  Tg or DTpg = Tp  Tg or in short DT), and Palancz (1974) for uniform droplets where the out-
the higher is the degree of stickiness. Because the Tg of let particle temperature was assumed to be the wet bulb
additive (glucose syrups or maltodextrin) is high, for temperature of air. The model was modified by Negiz
instance, Tg of maltodextrin is 205 C (Roos & Karel, et al. (1995) for non-uniform droplets, and the assump-
1991), then adding of additive increases the Tg of feed tion of wet bulb temperature for outlet particle tempera-
reducing the DT, which in turn lowers the stickiness ture was removed. This model is more economical than
behavior. Also, when cold air is introduced to the bot- the dynamic models in terms of time consumption for
tom of the dryer or the wall temperature is reduced, computing and was used in our study with some modifi-
the DT is decreased resulting in a lower stickiness. The cation. The first modification was the introduction of a
situation is similar for the case of controlling the outlet relative drying rate into the mass balance equation for
air temperature to a lower value. Since, Tg of particles is moisture content of the droplet that has been used by
a function of moisture content (Levine & Slade, 1988), different authors (Fyhr & Kemp, 1998; Langrish &
the drying performance becomes a factor affecting the Zbicinski, 1994; Strumillo & Kudra, 1986; Zbicinski,
DT or stickiness. In turn, the drying performance de- Grabowski, Strumillo, Kiraly, & Krzanowski, 1988;
pends on the inlet variables and the interaction between Zhang, Keey, Langrish, Pasley, & Kemp, 1994). The
the drying air and droplet properties. The use of DT to second modification was the application of three compo-
control the spray drying performance is a major concern nents of air and droplet velocities in the momentum
58 V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65

balance equations of droplets. Maltodextrin–sucrose 1993; Young, Munson, & Okiishi, 1997), h is the axial
(40/60) was used as a sugar-rich food model and is de- distance which is zero at the atomizer, dp is the droplet
noted by MDS46 in this paper. diameter. The following correlation for Cd is used:
24
2. Mathematical model Cd ¼ , Re 6 0:2 ð5Þ
Re

The model is developed for a conical based and 24 4


Cd ¼ þ , 0:2 < Re < 500 ð6Þ
co-current spray dryer using the following assumptions: Re Re1=3

• All the particles are assumed to move from the top to C d ¼ 0:44, 500 6 Re ð7Þ
the bottom part of the dryer chamber. qa U R d p
• Size of droplets after atomization follows a log–nor- Re ¼ ð8Þ
la
mal distribution.
• No agglomeration or break up of droplets or particles The radial distance (r) is then determined from Upr and
takes place in the dryer chamber. Upx (Eq. (9)), qa and la are the density and viscosity of
• The temperature of the droplets during drying is uni- air, respectively.
form, i.e., the temperature gradient within the drop- dr U pr
lets is negligible. ¼ ð9Þ
dh U px
• The droplet size is small enough that the calculation
of average moisture content is reasonable. The sur-
face moisture content is assumed to be in equilibrium 2.1.2. Mass balance for droplets (particles)
with the film layer of the air surrounding the droplets.
dmp Ap K m
This assumption is used to calculate the equilibrium ¼ n ðY e  Y Þ ð10Þ
humidity of the film layer. dh U px
• The air temperature and humidity along the cross- where, mp, n, Ap, Km, Ye and Y are the mass of particle,
section area of the dryer chamber are uniform. relative drying rate, droplet surface area, mass transfer
• There are no radial and tangential components of the coefficient, humidity at saturation and humidity at bulk
inlet drying air. The axial air velocity is assumed to be air, respectively. n = 1 for the constant rate period and is
uniform along the cross-section area of the dryer a function of moisture for the falling rate period as given
chamber. in Appendix A. Km can be calculated by Eq. (11).
K m ¼ qa Dv Sh=d p ð11Þ

2.1. Equations of the model Dv is diffusivity of water in air calculated by the Fuller
formula (Perry & Green, 1997), Sh is the Sherwood
2.1.1. The trajectory of droplets (particles) number. Ye is taken at film temperature (Tfilm) of the
Let Up and Ua be the velocity of particles and air, air close to particle surface, where Tfilm is taken as par-
respectively. The subscripts x, r and t represent the axial, ticle temperature (Tp).
radial and tangential components.
" ! #
dU px qa 3 qa C d U R ðU px  U ax Þ 1 2.1.3. Heat balance for droplets (particles)
¼ 1 g ð1Þ
dh qp 4 qp d p U px dT p
¼ ½pd p K a NuðT a  T p Þ
" # dh
dU pr U 2pt 3 qa C d U R ðU pr  U ar Þ 1 1
¼  ð2Þ þ ðH f g þ C pv ðT a  T p ÞÞ ð12Þ
dh r 4 qp d p U px ms ðC ps þ XC pw ÞU px
" # where Tp is the particle temperature, Ta is the dry bulb
dU pt U pt U pr 3 qa C d U R ðU pt  U at Þ 1 temperature of the bulk air in the control volume, Hfg
¼   ð3Þ is the latent heat of water evaporation in the product,
dh r 4 qp d p U px
which is given in Appendix A. Cpv, Cps and Cpw are
where UR is the relative velocity between the droplet and the specific heats of vapor, solid and water, respectively.
the air (Zbicinski, 1995), X and ms are the moisture content (w/w dry basis) and
2 2 2 1=2
U R ¼ ½ðU px  U ax Þ þ ðU pr  U ar Þ þ ðU pt  U at Þ  dry matter of the particle, mp = Xms. Nu is the Nusselt
number.
ð4Þ
The critical moisture contents of maltodextrin and
and Cd is the drag force coefficient that can be deter- sucrose and their mixtures were above 2.7 as reported
mined from Reynolds number (Koo & Kuhlman, by Truong (2003). However, in this modelling work,
V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65 59

the initial moisture content of feed was below 2, i.e., the 2.1.5. Heat balance for drying medium
drying took place in the falling rate period only. Some  
dH h X 1 dmp K a pd p NuðT a  T p Þ
authors (Negiz et al., 1995) used the critical moisture ¼ Hv  ð19Þ
dh droplets G dh U px
content as the value where the particle size stoped
shrinking. In the case of drying taking place in the fall- H h ¼ C pa ðT a  273:16Þ þ YH v ð20Þ
ing period only, the droplet size is expected to change
due to shrinkage. For simplification, the droplet diame- H v ¼ C pv ðT a  T dewp Þ þ hf gdewp þ C pw ðT dewp  273:16Þ
ter dp is updated based on the assumption of balloon
shrinkage without crust forming (Eq. (13)), where dpi ð21Þ
is the initial droplet diameter, qpi and qp are the initial where Hh and Hv are the enthalpies of humid air and
density and corresponding density at dp of the droplet: vapor in air, respectively (ASAE, 1994/1997; Brooker,
!1=3 Bakker-Arkema, & Hall, 1992). Tdewp and hfgdewp are
qpi  1000 the dew point temperature and latent heat of vaporiza-
d p ¼ d pi ð13Þ
qp  1000 tion at dew point temperature, respectively.

1þX 2.1.6. Glass transition temperature of particles


qp ¼ q ð14Þ The glass transition temperature of particles at any
1 þ X qqws s
moisture content X is a function of three components
where qs and qw (1000 kg/m3) are the density of solid of MD, sucrose, and water fractions which can be pre-
and water, respectively. For mixture MD–sucrose, the dicted by the analytical equation developed by Truong,
weighted average method was used for solid density Bhandari, Howes, & Adhikari (2002) (Eq. (22)).
(qMD = 1400 kg/m3 and qsuc = 1526 kg/m3, Truong, K MS xS ð1  xW ÞT gS þ K SW xW T gW
2003). T gMSW ¼
K MS xS þ xM 1  xW þ K SW xW
qs ¼ xMD qMD þ ð1  xMD Þqsuc ð15Þ xM ð1  xW ÞT gM þ K MW xW T gW
þ
K MS xS þ xM 1  xW þ K MW xW
The correlation for heat and mass transfer coefficients
ð22Þ
from the surface of droplets to the surrounding air,
Sherwood number (Sh) and Nusselt number (Nu), are where TgM (205 C), TgS (65 C), and TgW (135 C)
calculated using Ranz and Marshall equations (Ranz (Franks, 1985; Levine & Slade, 1988) are the glass tran-
& Marshall, 1952) for pure liquids. sition temperature values for maltodextrin DE6,
sucrose, and water, respectively. KMS (3.37), KMW (7.7)
Sh ¼ 2 þ 0:6Re0:5 Sc1=3 ð16Þ (Roos & Karel, 1991), and KSW (4.42) (Truong et al.,
2002) are the constants in the Gordon–Taylor equation
Nu ¼ 2 þ 0:6Re0:5 Pr1=3 ð17Þ (Gordon & Taylor, 1952) for binary mixtures of MD–
sucrose, MD–water, and sucrose–water, respectively.
where Sc = la/(qaDv) and Pr = Cphla/Ka are the Schmidt
and Prandlt numbers. la, Cph and Ka are the viscosity, xM, xS, and xW are the fractions of MD, sucrose, and
water in the particles. The above fractions are deter-
specific heat and thermal conductivity of humid air,
mined as below for a known ratio between solid frac-
respectively.
tions RMS (MD/sucrose ratio),
For a step size of dh, after calculation of the above
equations for each fraction of the droplet diameter X
xW ¼ ð23Þ
(from log–normal distribution), the drying process will 1þX
be completed by a balance of heat and mass for the dry-
ing air in the control volume (cross-section area of the 1  xW
xS ¼ ð24Þ
dryer chamber multiplied by dh) as shown in Eqs. (18) 1 þ RMS
and (19).
xM ¼ RMS xS ð25Þ
2.1.4. Mass balance for drying medium The calculation of the drying process of the droplets
from atomization to the outlet of the chamber was car-
 
dY b X dmp ried out in association to the calculation of the temper-
G ¼  ð18Þ ature difference (DTpg) between the particle temperature
dh droplets dh
and its glass transition temperature.
G is the mass flow rate of dry air, Yb is the humidity of
the bulk air in the control volume. The right-hand side is 2.1.7. Overall moisture content and temperature difference
the total of evaporated water of all the droplets in the When all the fractions (j) of particles reach the
control volume. bottom wall of the dryer, they are considered to be
60 V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65

homogeneously mixed together and the overall moisture ied from 1 to 3 h on PC computer (56 MB RAM, AMD
content Xov, and the difference of particle temperature 3D processor) depending on the mean size of the drop-
and its glass transition temperature (DTov) are calcu- lets. Related information for drying modelling is shown
lated as below: in Appendix A.
P
droplets F ðjÞX j msj
X ov ¼ P ð26Þ
droplets F ðjÞmsj 3. Validation of the model
X
DT ov ¼ F v ðjÞDT j ð27Þ 3.1. Experiment
droplets

where F(j) and Fv(j) are the frequencies of the number The model was validated by spray drying experiments
and volume distributions of the droplet size of fraction of a maltodextrin–sucrose mixture (4:6). The setup and
j. msj is the solid mass of the droplet of fraction j. The method for measurement of DTpg between wall temper-
numerator and denominator in Eq. (26) are the total ature and the glass transition temperature of the depos-
water amount in the particles and the dry mass of the ited layer are described in Appendix B. The inlet
particles, respectively. These values were used to com- parameters for drying are shown in Table 1.
pare with the experimental measurement of dried Run 0 was conducted to compare between the pre-
particles. dicted and measured air temperature profile along the
vertical distance inside the dryer chamber. Runs 1–4
without cooling air were conducted to compare the pre-
2.2. Solution of the model dicted and measured DTpg. Inserting of cooling air to the
dryer is described in Truong (2003). Runs 5 and 6 were
The model of co-current spray drying is a set of 8 or- carried out to compare outlet air temperature between
dinary differential equations including Eqs. (1)–(3), (9), calculated and measured values. The recovery of pow-
(10), (12), (18) and (19). This set of equations was solved ders in the cyclone and from different parts of the dryer
using a fourth-order Runge–Kutta procedure. For each walls was also compared. Two replicates were carried
step size dh, firstly the heat and mass balance and the out for all experiments.
trajectory were calculated for each droplet and then Determination of size distribution of the droplets
the change in humidity and enthalpy of the control vol- generated by an atomizer can be either directly by mea-
ume were contributed by the fraction of that drop size suring the droplets below the atomizer as described else-
(total of the drops in fraction of corresponding size). Be- where in the literatures or indirectly by inferring from
cause the initial vertical velocity (Upx) of the droplet is measured data on the dried particles. The purpose of
equal to 0 (at t = 0), an initial axial velocity of 0.1 m/s this section is to determine the size distribution of the
was set for all the fractions in order to start the calcula- droplets as well as of the dried particles from the
tion. It was found that the initial axial velocity between pilot-scale spray dryer.
0.05 and 0.4 m/s provided the same final results. The The size distribution of the droplets was determined
droplets/particles reach the terminal velocity rapidly from distribution data of the dried particles. Spray-dried
depending on the drop size. Because the velocity of the powder of MDS46 was obtained from spray drying runs
particles is not the same, the largest particles are as- at 145 C for 5 min and atomizer speeds of 20,000,
sumed to contribute the moisture firstly into the control 18,000, and 16,000 RPM. It is assumed that there is no
volume and next for the smaller particles. The evapo-
rated moisture from the particles added to the control
volume followed the above order until all the fractions Table 1
are used. The total added moisture from all the fractions Inlet parameters for spray drying experiments of maltodextrin–sucrose
after each step dh increases the humidity of the control mixture (M:S = 4:6) with feed concentration of 50%
volume. The procedure is same for enthalpy and temper- Run Drying air Air flow Feed Cooling air Speed of
ature of air of the control volume. After all the fractions number temperature, rate, kg/s rate, fraction, % atomizer
contributed the moisture and enthalpy to the control C kg/s
volume, the droplet size was updated by Eq. (13). The 0 145 AF1 F0 0 18,000
outlet parameters from this step become the inlet param- 1 135 AF1 F1 0 20,000
2 155 AF1 F1 0 20,000
eters for the next step and so on. The program stops
3 135 AF2 F1 0 20,000
when all the fractions reach the exit of the dryer. The 4 155 AF2 F1 0 20,000
distance from the atomizer to the exit was taken as 5 135 AF3 F1 0 20,000
1.6 m. The program was written by Qbasic language. 6 135 AF3 F1 12 20,000
Initially, the program was started by a step size Note: AF1 = 0.03248, AF2 = 0.03794, AF3 = 0.03690 (kg dry air/s).
dh = 104 m. The time for running of the program var- F0 = 0.000545, F1 = 0.000708 (kg/s). Speed of atomizer is in RPM.
V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65 61

caking during that short time of drying and the particles


reach the final size. The volume distribution of dried
particles was analyzed by the laser diffraction method
using a Malvern Mastersizer. The information on dried
particles was used to determine the size distribution of
the droplets.

3.2. Droplet and particle size distribution

The prediction and measurement of size distribution


of the particles is shown in Fig. 1. The graph shows that
the size distribution generated for the droplets from
atomizer based on Sauter-mean diameter calculated by Fig. 2. Comparison between predicted and measured air and wall
the Scott formula (A.4) and an empirical standard devi- temperature profile in spray drying of MDS46 for the drying condition
ation relationship (Eq. (A.1)) in association with a of run 0 in Table 2.
shrinkage ratio of 1.25 provided a fairly close shape of
predicted dried-particle distribution compared to mea-
surement.

3.3. Air and wall temperature

The calculated air temperature profile was about 2–


4 C below the measured wall and air temperature in
the cylindrical chamber but almost the same as those
in the conical bottom chamber for the drying condition
of run 0 (Fig. 2). The wall temperature was slightly
lower than the air temperature. The temperature in the
conical section was nearly constant against axial
distance.

3.4. Moisture and glass transition temperature profile

The prediction of temperature difference DTpg


between particle temperature and its Tg, the glass transi-
tion temperature, and moisture content of particle of Fig. 3. Plot of temperature difference DT(Tp  Tg) of particles during
different size fractions as a function of axial distance spray drying process of MDS46 for the drying condition of run 4 in
Table 1.

Fig. 1. Size distribution of the droplets and dried particles in spray Fig. 4. Plot of particle glass transition temperature during drying
drying of MDS46 for the drying condition of run 0 in Table 1. process of MDS46 for the drying condition of run 4 in Table 1.
62 V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65

Table 2
Comparison between calculated and measured temperature difference
and moisture content of powder obtained by spray drying of MDS46
Run Calculated Measured Moisture content of
no. DTpg, C DTpg, C powder at cyclone, % db
Overall Cyclone Measured Calculated
1 12.9 8.8 7.3–13.2 2.5 2.3
2 23.6 22.2 16.1–18.8 2.7 1.7
3 16.5 13.9 8.0–12.9 2.2 2.0
4 32.8 32.3 18.5–29.6 1.9 1.4
Note: Drying condition of the runs in this table is corresponding to
that of the run number in Table 1.

(DTcyclone) as well as overall moisture content (Eq.


(26)) were compared to measured data from different in-
let drying conditions as shown in Table 2. The calcu-
Fig. 5. Plot of particle moisture content during drying process of
MDS46 for the drying condition of run 4 in Table 1.
lated DTpg (or DTcyclone) values are the average
temperature difference of all the fractions (or fractions
except semi-wet particles) when the particles reach the
from atomizer for drying condition of run 0 are given in bottom wall. The measured temperature difference var-
Figs. 3–5. ied in a wide range. The setup for this determination is
The results show that the drying process takes place described in Appendix B. Initially, the temperature dif-
quickly for the initial droplet size of less than 70 lm, ference was high due to low Tg (high moisture). After
i.e., the drying time is less than 0.1 s for droplets of 30 min, the powder in the bottom part of the chamber
below 30 lm and less than 1 s for droplet size of was dried further increasing its Tg. Therefore, the tem-
70 lm. Due to the fast drying, the glass transition tem- perature difference decreased with operation time. This
perature of the particles of the initial droplet size below interaction can be described by consideration of the con-
100 lm reaches 60 C at a short distance of 0.2 m from ductive drying of the particles in contact with the cham-
the atomizer (Fig. 4). Moisture content of the major par- ber wall presented in Truong (2003). Table 2 shows that
ticles (initial droplet size less than 110 lm) decreases to calculated temperature differences of runs 1 and 3 are
below 0.04 (db) before moving to the conical bottom close to measured data. However, the calculated values
chamber (Fig. 5). The drying time occurs too fast for were higher than measured data for runs 2 and 4. The
the molecules to rearrange. Thus, the solution maintains temperature difference calculated for the powder in the
an amorphous form at any stage of the drying process. cyclone was closer to measured values than that calcu-
This argument is confirmed by the fact that the final lated for all the fractions of powder. Moisture content
product is completely amorphous as no melting peak of powder in the cyclone was similar for prediction
was observed in DSC thermogram (Truong, 2003). In (Eq. (26)) and measurement except for run 2. The mea-
other words, the calculation of the glass transition tem- sured moisture content of run 2 was high (2.7%) proba-
perature of the droplet at any stage of the drying process bly due to error during handling of the sample. This
is valid. moisture content should be lower than that of run 1
For atomizer speeds of less than 20,000 RPM with (2.5%) because the drying temperature is higher.
the same geometry of the dryer, droplets with size below
100 lm do not hit the dryer wall under centrifugal force 3.5. Particle temperature profile
for the feed concentration range of 40–50% (w/w) as cal-
culated from the model. These fractions occupy about The temperature profile of air and particles are given
90% of total volume of feed and level off the moisture in Fig. 6. The small particles approach the drying air
content and temperature difference before moving to temperature rapidly. Whereas, the large particles ap-
the conical bottom chamber (Figs. 3–5). Running the proach wet bulb temperature rapidly and then slowly in-
model for feed rates in the range of 0.000708– crease to a lower temperature compared to outlet air
0.001188 kg/s, initial moisture content of feed of 1–1.5, temperature. For the droplets with size less than
and the speed of the atomizer of 16,000–20,000 RPM, 100 lm, particle temperature approached air tempera-
the amount of large particles hitting the dryer wall under ture at a distance of 0.2 m from atomizer. Temperature
centrifugal force was from 6% to 12% by weight of the of small particles with initial droplet size of below 40 lm
feed. Thus, the deposition of the semi-wet particles is reached a temperature of above 100 C shortly and then
at least in this range. decreased together with air temperature. The excess of
The calculated temperature difference (Eq. (27)) temperature of small particles over 100 C is agreed with
for all the fractions (DTpg) and for cyclone powder calculation of Zbicinski et al. (1988).
V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65 63

A.3. Droplet size distribution

The results (Truong, 2003) showed that the stan-


dard deviation of particle size was proportional to the
Sauter-mean diameter of the dried particles (Dvsd),
R2 = 0.994.
ln rgd ¼ 0:01396Dvsd þ 1:07753 ðA:3Þ
Because the formation of the skin surrounding the par-
ticle is neglected, the Sauter-mean of the wet particles or
droplets (Dvsw) is proportional to that of the dried par-
ticles. For the small scale atomizer, the Scott formula is
Fig. 6. Drying air (s) and particles (line) temperature profile as a suitable for prediction of the Sauter-mean of the drop-
function of axial distance for the drying condition of run 5 in Table 1. lets created by the atomizer as shown in the following
equation (Masters, 1994):
4. Conclusion
0:171 0:537 0:017
Dvs ¼ 5904:7ðM p Þ ðpd a N a Þ ðlF Þ ðA:4Þ
A mathematical model has been developed to de-
scribe the stickiness in spray drying of sugar-rich foods where Mp = QF/nv/hv is the mass flow per total periph-
based on the glass transition temperature approach. ery, da(m), nv, hv (m), and Na (RPM) are the diameter
The model of cocurrent spray drying is a steady state of the disc, number of vanes in the disc, the height of
one-dimensional one with log–normal distribution of the vanes and the speed of rotation of the atomizer,
droplets generated by the rotary atomizer. The differ- respectively. QF (kg/s) and lF (Pa s) are the feed rate
ence between particle temperature and its glass transi- and viscosity of feed. The Sauter-mean is expressed in
tion temperature (DTpg) was calculated for all particle lm units. The shrinkage ratio Ksh = Dvsw/Dvsd was
sizes as a function of distance from atomizer. The pre- found to be 1.25 ± 0.005 for the three runs. Therefore,
diction and measurement of outlet air temperature Eqs. (A.5)–(A.9) below were used to generate the distri-
agreed well for different drying conditions. bution of the droplets from atomizer:
The model can be used to minimize the stickiness by ln rgw ¼ 0:01396Dvsw =K sh þ 1:07753 ðA:5Þ
selecting of the inlet drying variables to minimize the
DTpg at the exit of the dryer as done in this work and ln DgVw ¼ ln Dvsw þ 0:5ðln rgw Þ
2
ðA:6Þ
presented in Truong et al. (2004). Expansion to other
food models is also possible by using physical proper- The geometric mean of the number distribution DgNw
ties, drying kinetics, sorption isotherm and size distribu- can be determined by Eq. (A.7) given below (Allen,
tion data of such foods in the drying model. 1997):
2
ln DgNw ¼ ln Dvsw  2:5ðln rgw Þ ðA:7Þ
Appendix A. Related information for modelling
where DgVw and rgw are the geometric mean of volume
distribution and standard deviation of the droplets,
A.1. Sorption isotherm for MDS46 using modified
respectively. The frequency of number distribution
Henderson equation
F(lndp), then known as density function f(lndp), can be
determined from DgNw and rgw using Eq. (A.8):
lnð1  RHÞ ¼ c1 ðT þ c2 Þð100X Þ
m
ðA:1Þ " #
2
1 ðln d p  ln DgNw Þ
where RH is given as a fraction, T is in Kelvin, and X is f ðln d p Þ ¼ pffiffiffiffiffiffi exp  ðA:8Þ
ln rgw 2p 2ln2 rgw
moisture fraction, dry basis. The constants c1, c2, and m
are 0.000405, 187.962, and 1.169, respectively. F ðjÞ ¼ F ðln d p Þ ¼ f ðln d p Þd ln d p ðA:9Þ

A.2. Latent heat of evaporation of MDS46 (J/kg) The number of droplets Ndrop generated by the atomizer
can also be obtained, where QF is the feed rate of the
solution,
H f g ¼ A1 þ B1 T þ ðA2 þ B2 T Þ expðA3 X Þ ðA:2Þ
QF
N drop ¼ P ðA:10Þ
where T is air temperature (K) and X is product mois- pd 3p
qpi droplets F ðln d p Þ 6
ture content (db). The constants are A1 = 3.1613 · 106,
B1 = 2413.7, A2 = 232,582, B2 = 179.6, and A3 = where qpi is the density of feed which is equivalent to the
5.089. initial density of the droplets.
64 V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65

Table A.1 1 Heater


Parameters used for drying kinetics of MD, sucrose and their mixtures
Ta, C Xc X0 a1 x0/xc Fan
200mm 250mm
Sucrose P135 4 1.5 0 1
<135 4 1.5 0.003Ta + 0.4056 1 + a1(X0  Xc) 500mm
Points for temperature
250mm measurement of:
MD 2.7 2.7 1/(Xc  Xe) 1
1. inlet air
Note: MD is for maltodextrin. 2. outlet air
__ wall
250mm

Aluminium tape
A.4. Relative drying rate (n) 1600mm
300mm Thermocouple tip
Wall
By taking the ratio x/xc between the drying rates in
the falling and constant rate periods, the general model
for n = x/xc was developed for maltodextrin and Teflon layer
500mm Note: Wall temperature
sucrose: measurement technique
9
X P Xc : n¼1 >
> 2
>
>
X 0 6 X 6 X c : n ¼ 1  a1 ðX c  X Þ >
>
=
X 6 X0 : n ¼ x0 =xc þ a2 ðX  X 0 Þ Fig. B.1. Spray drying experimental setup.
>
>
¼ x0 =xc ðX  X e Þ=ðX 0  X e Þ >
>
>
>
;
x0 =xc ¼ 1 þ a1 ðX 0  X c Þ
ðA:11Þ Measurement of the outlet air temperature used an
adaptation of the method described by Papadakis &
where a1 and a2 are the slopes of the two straight lines. King (1988a) where a Teflon cap of peaked shape was
x0 and xc are the drying rates at X0 and Xc, respectively. placed above the thermocouple tip to prevent contact
Xe is the equilibrium moisture content calculated from between the particles and the tip. A similar technique
the sorption isotherm in Eq. (A.1) (Xe 6 X0 6 Xc). was applied for the inlet air temperature thermocouple
These values are shown in Table A.1 where Ta is the dry- located above the atomizer but the Teflon cap was
ing air temperature. The relative drying rate of mixtures replaced by a vertical Teflon cylinder to avoid contact
is assumed to follow the weighted average method, between the droplets and the tip.
where xMD is the fraction of MD in the solid content, The measured DTpg was considered to be the differ-
w/w. ence between outlet air temperature and the Tg of prod-
nmix ¼ nSuc ð1  xMD Þ þ nMD xMD ðA:12Þ ucts. The Tg of products was taken as powders at
cyclone and bottom chamber after drying. Therefore,
The n was not significantly different among tempera-
the measured DTpg was varied in a wide range as shown
tures for maltodextrin but significantly different for
in Table 2.
sucrose and therefore MDS46 (Truong, 2003).

References
Appendix B. Experimental setup for measurement of wall
temperature and air temperature during spray drying Allen, T. (1997). Powder sampling and particle size measurement (5th
process of maltodextrin–sucrose in pilot dryer ed., Vol. 1). London: Chapman and Hall.
ASAE, S. (1994/1997). Engineering practices and data, Psychrometric
data.
A pilot-scale Anhydro spray dryer (Lab1) with a ro- Bhandari, B. R., Datta, N., Crooks, R., Howes, T., & Rigby, S. (1997).
tary atomizer was used in this research. The diameter A semi-empirical approach to optimise the quantity of drying aids
of the drying chamber is 1 m. Fig. B.1 shows the setup required to spray dry sugar-rich foods. Drying Technology, 15,
for measurement of the wall and air temperature of 2509–2525.
Bhandari, B. R., Datta, N., & Howes, T. (1997). Problems associated
the dryer chamber. An insulated T-type copper– with spray drying of sugar-rich foods. Drying Technology, 15,
constantan thermocouple was used (0.51 mm wire diam- 671–684.
eter). For wall temperature measurement, the tip of Bhandari, B. R., Senoussi, A., Dumoulin, E. D., & Lebert, A. (1993).
thermocouples were covered by Teflon tape with a thick- Spray drying of concentrated fruit juices. Drying Technology, 11,
ness of 0.5 mm. This Teflon layer was applied to avoid 1081–1092.
Brennan, J. G., Herrera, J., & Jowitt, R. (1971). A study of some of the
direct contact between the tip and particles. Aluminium factors affecting the spray drying of concentrated orange juice, on a
tape was used to fix the thermocouples on the dryer laboratory scale. Journal of Food Technology, 6(September),
walls. 295–307.
V. Truong et al. / Journal of Food Engineering 71 (2005) 55–65 65

Brooker, D., Bakker-Arkema, F., & Hall, C. (1992). Drying and Papadakis, S. E., & King, C. J. (1988b). Air temperature and humidity
storage of grains and oilseeds. New York: Van Nostrand Reinhold. profiles in spray drying. 2. Experimental measurements. Industrial
Clement, K. H., Hallstrom, A., Dich, H. C., Lee, C. M., Mortensen, J., and Engineering Chemistry Research, 27, 2116–2123.
& Thomsen, H. A. (1991). On the dynamic behavior of spray Parti, M., & Palancz, B. (1974). Mathematical model for spray drying.
dryers. Transaction of Institution of Chemical Engineering, 69, Chemical Engineering Science, 29, 355–362.
245–252. Perry, R. H., & Green, D. W. (1997). PerryÕs chemical engineering
Crowe, C. T. (1980). Modelling spray–air contact in spray drying handbook (7th ed.). New York: McGraw-Hill.
systems. In A. S. Mujumdar (Ed.), Advances in drying (pp. 63–99). Ranz, W. E., & Marshall, W. R. J. (1952). Evaporation from drops.
Washington: Hemisphere Publishing Corporation. Chemical Engineering Progress, 148, 141–146.
Crowe, C. T., Sharma, M. P., & Stock, D. E. (1977). The particle- Roos, Y., & Karel, M. (1991). Water and molecular weight effects on
source-in-cell (PSI-CELL) model for gas-droplet flows. Journal of glass transitions in amorphous carbohydrates and carbohydrate
Fluid Engineering, 9, 325–332. solutions. Journal of Food Science, 56, 1676–1681.
Dickinson, D. R. M., & Marshall, W. R. J. (1968). The rates of Strumillo, C., & Kudra, T. (1986). Drying: Principles, applications and
evaporation of sprays. AIChE Journal, 14, 541–552. design. London: Gordon and Breach Science Publishers.
Franks, F. (1985). Biophysics and biochemistry at low temperatures. Truong, V. (1994). Spray drying of tamarind concentrate and its quality
Cambridge: Cambridge University Press. evaluation. Master Thesis. Bangkok, Thailand: Asian Institute of
Fyhr, C. K., & Kemp, I. C. (1998). Evaluation of the thin-layer Technology.
method used for measuring single particle drying kinetics. Trans- Truong, V. (2003). Modelling of the glass transition temperature of
actions of IChemE, 76, 815–822. sugar-rich foods and its relation to spray drying of such products.
Gordon, M., & Taylor, J. S. (1952). Ideal copolimers and the second- Ph.D. Dissertation. Australia: University of Queensland.
order transitions of synthetic rubbers. I. Non-crystalline copoli- Truong, V., Bhandari, B. R., & Howes, T. (2004). Optimization of
mers. Journal of Applied Chemistry, 2, 493–500. cocurrent spray drying process of sugar rich foods. Part II—
Gupta, A. S. (1975). Spray drying of orange juice. US patent No. Optimization of spray drying process based on glass transition
4112130. concept. Journal of Food Engineering, in press, doi: 10.1016/
Koo, Y. M. K., & Kuhlman, D. K. (1993). A variable flow nozzle with j.jfoodeng.2004.10.018.
consistent spray performance. Transaction of ASAE, 36, 685–690. Truong, V., Bhandari, B. R., Howes, T., & Adhikari, B. (2002).
Langrish, T. A. G., & Zbicinski, I. (1994). The effects of air inlet Analytical model for the prediction of glass transition temperature
geometry and spray cone angle on the wall deposition rate in spray of food systems. In H. Levine (Ed.), Amorphous food and
dryers. Transaction of IChemE, 72, 420–430. pharmaceutical systems (pp. 31–47). The Royal Society of
Lazar, W. E., Brown, A. H., Smith, G. S., Wong, F. F., & Lindquist, Chemistry.
F. E. (1956). Experimental production of tomato powder by spray Young, D. F., Munson, B. R., & Okiishi, T. H. (1997). A brief
drying. Food Technology, 129–134. introduction to fluid mechanics. New York: John Wiley.
Levine, H., & Slade, L. (1988). Water as a plasticizer: Physical– Zaleski, J., Lipowska, T., & Kuszlik, J. (1968). Spray drying of
chemical aspects of low-moisture polymeric systems. In F. Franks concentrated fruit juices. Prace Instytutow I Laboratoriow Bad-
(Ed.), Water science reviews 3 (pp. 79–185). Cambridge: Cambridge awczych Przemysluspozywczego, 18(4), 53.
University Press. Zbicinski, I. (1995). Development and experimental verification of
Masters, K. (1994). Spray drying handbook. London: Longman momentum, heat and mass transfer model in spray drying. The
Scientific & Technical. Chemical Engineering Journal, 58, 123–133.
Negiz, A., Lagergren, E. S., & Cinar, A. (1995). Mathematical models Zbicinski, I., Grabowski, S., Strumillo, C., Kiraly, L., & Krzanowski,
of cocurrent spray drying. Industrial and Engineering Chemistry W. (1988). Mathematical modelling of spray drying. Computers and
Research, 34, 3289–3302. Chemical Engineering, 12, 209–214.
Oakley, D. E., & Bahu, R. E. (1993). Computational modelling of Zhang, G. J., Keey, R. B., Langrish, T. A. G., Pasley, H. S., & Kemp,
spray dryers. Computers and Chemical Engineering, 17, S493–S498. I. C. (1994). An experimental test of the concept of the character-
Papadakis, S. E., & King, C. J. (1988a). Air temperature and humidity istic drying curve using the thin-layer method. In: V. Rudolph, &
profiles in spray drying. 1. Features predicted by the particle source R. B. Keey (Eds.). Drying 94: Proceedings of the 9th International
in cell model. Industrial and Engineering Chemistry Research, 27, Drying Symposium, Gold Coast, Australia (pp. 123–130).
2111–2116.

You might also like