You are on page 1of 12

Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

Contents lists available at ScienceDirect

Food and Bioproducts Processing

journal homepage: www.elsevier.com/locate/fbp

Modeling of food drying processes in industrial


spray dryers

Hugo M. Lisboa ∗ , Maria Elita Duarte, Mário Eduardo Cavalcanti-Mata


Unidade Academica de Engenharia de Alimentos, Universidade Federal de Campina Grande, Av. Aprigio Veloso 882,
58429-200 Campina Grande, Paraíba, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: The shift from a trial-and-error approach on food product development to a quality-by-
Received 15 June 2017 design paradigm requires tools that support the scientist in making decisions for the design
Received in revised form 23 August of research and development activities. Presently, many of these tools require high invest-
2017 ments in software or time for implementation. Consequently, the present work had the
Accepted 18 September 2017 objective to develop a simplified model of the spray drying process in an industrial spray
Available online 3 November 2017 dryer to support activities of food product development and that can be easily implemented
in any software. The model was verified and validated using an industrial spray dryer in
Keywords: food drying processes covering a wide range of operating conditions and food products.
Food products The model was further extended so that the scientist could estimate not only key operating
Heat-Mass balances parameters such as feed flow rate or drying gas outlet temperature but also final particle size
Drying kinetics and drying kinetics for a better understanding of the drying phenomenon by spray drying.
Droplet size From the comparison of the experimental and the estimated results it is concluded that
Modeling the model successfully describes all spray drying operations independently of the product
Scale-up nature. Raw material and laboratory time can be reduced by replacing the traditional trial
and error methodologies by using the developed tool. Estimates of wet bulb temperature,
dew point and glass transition temperature increase process knowledge preventing process
errors. Scale-up is also facilitated by the use of non-dimensional estimated parameters such
as outlet relative humidity.
© 2017 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction processes due to the high number of experiments required (Musina


et al., 2017). Consequently several researchers turn to approaches that
The development of new processes is often a time-consuming and use more fundamental tools requiring greater knowledge on dynam-
costly process, since most of the times they are conducted through trial ics of the processes as well as the equipment itself (Baldinger et al.,
and error methodologies, making them complex and extremely long 2012; Ivey and Vehring, 2010). A fundamental or mechanistic approach

Abbreviations: DOE, design of experiments; CFD, Computational Fluid Dynamics; Qin , heat entering spray drying; Qfeed , heat required to
dry the feed; Qout , heat coming out of the equipment; Qloss , heat lost to the exterior; Fdrying , drying gas flow rate; Cp,x , specific heat; Tin , inlet
temperature drying gas; Tout , outlet temperature drying gas; Tfeed , feed initial temperature; Cfeed , feed initial concentration; Ffeed , feed flow
rate; Hvap , vaporization enthalpy; Twb , wet bulb temperature; U, overall heat transfer coefficient value; Asup , equipment superficial area;
Tw , logarithmic mean of temperatures; mi,j , mass of water; Hin , drying gas moisture content inlet; Hout , drying gas moisture content
outlet; Pvout , vapor pressure at outlet; Pvsat , vapor pressure saturation; % RHout , relative humidity at outlet; Tdp , dew point temperature;
Tg, glass transition temperature; N, number of blades; D, disk diameter; ω, blade speed; feed , feed viscosity;  feed , feed superficial tension;
Pfeed , feed differential pressure; feed , kinematic viscosity of feed; gas , kinematic viscosity of drying gas; Fatom , Atomization gas flow rate;
vatom , Atomization gas velocity; D50, mean droplet size; Ddroplet , mean droplet size; Dparticle , mean particle size;  SD , standard deviation;
Nu, Nusselt number.

Corresponding author.
E-mail address: hugolisboa.certbio@ufcg.edu.br (H.M. Lisboa).
https://doi.org/10.1016/j.fbp.2017.09.006
0960-3085/© 2017 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
50 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

to a process will need to involve models that can be generally appli- of atmospheric air at 25 ◦ C), the electric heating resistance has
cable to any product and help defining operative conditions to obtain the capacity to heat the drying gas up to 250 ◦ C. Additionally,
a product within certain quality attributes (Nath and Satpathy, 1998). the surface area of the equipment is estimated to be 11.1 m2 .
Additionally, the fundamental approach to a process will allow a bet- The equipment has a cyclone coupled for dust separation and
ter understanding about the most critical factors to obtain specific
operates on open cycle, i.e. the drying gas is not reused. The
properties in a product. Considering the numerous inputs and out-
equipment control system allows the set-point to be set at the
puts, one of the processes that can be significantly benefited from such
drying gas inlet temperature or at the outlet temperature.
approach is the spray drying operation since it is a complex product
preparation process. Even though the basic principles of spray drying
may seem simple, it can become very complex due to the choice of 2.2. Spray drying basic model
operational parameters, atmospheric conditions, food product prepa-
ration and formulation. In common practice, the development of the Spray drying is a well-known and mature operation capable
spray drying process is often empirical and is conducted experimen- of transforming solutions, suspensions or emulsions into a
tally. Traditional methods use a design of experiments (DOE) (Maltesen solid product. The spray drying process can be defined as an
et al., 2008) or statistical treatment of process parameters and product operation, wherein a liquid stream pumped into an atomizer
properties (Prinn et al., 2002). This is often a time consuming exercise,
is constantly divided into very fine droplets within the dry-
requiring large amounts of raw material, and the resulting process is
ing chamber. Here the fine droplets contact a hot gas, which
often not well understood or sufficiently robust. Recent efforts have
by convection provides energy to heat and vaporize most of
focused on the application of a spectrum of fundamental models for the
understanding of the spray drying process, ranging from steady state the solvent present in the droplets thereby forming powder
and equilibrium approaches to Computational Fluid Dynamics (CFD) particles, which are separated from the drying gas using a
(Oakley, 2004). However, many published papers on spray drying mod- cyclone or a bag filter. In the development of the spray dry-
els are more focused on complex fluid dynamics computation methods ing model it was considered that the geometry of the drying
or dedicated to the development of pharmaceuticals rather than food chamber was constituted by a cylinder and cone not consid-
products (Grasmeijer et al., 2013; Woo and Bhandari, 2013). Other mod- ering the flow patterns of the drying gas and simply assuming
els are merely applicable to laboratory scale equipment and operating that it is continuous inside that geometry and ideally mixed.
modes other than the operation of an industrial spray dryer in the food
Thus the parameters that differentiate the operation between
area. Other models are especially dedicated to the drying of a certain
one equipment and another are concentrated on the heat loss
product as in the case of milk (Silva et al., 2017). Recent advances in
along the walls of the equipment.
the modeling of spray drying by CFD have been discussed, including
its use in demanding industries such as the pharmaceutical industry
(Fletcher et al., 2006). The complexity of the spray drying process, where 2.2.1. Heat balance
phenomena occur on vastly different time and spatial scales, suggests In a first stage we will focus on the development of the heat
that the computational power still needs to evolve substantially before balance to make estimates of the drying gas flow rate, the feed
the particle formation process can be adequately predicted within a flow rate, the drying gas inlet temperature and the drying gas
plant-scale CFD simulation. More importantly, many created models outlet temperature. Fig. 1 summarizes the position where each
lack validation. Considering food engineering works using spray dry- equation is considered. Thus, the generic energy balance of
ing, the information provided by the existing papers on models is barely
the spray dryer is given by:
used, and the reason is because such models are difficult to use and
the scientist is not really involved in the creation and consequently it
Qin = Qfeed + Qout + Qloss (1)
is hard to grasp the outcomes of such models. Hence, there is a gap
in the creation of a simple, precise and robust model that allows to
estimate drying conditions on industrial spray dryers, and that can be where Qin is the heat entering the equipment by means of
easily implemented using spreadsheets or other code free software. the drying gas, Qfeed is the heat spent in heating the product
The industrial application of this type of model would serve to bet- droplets to the wet bulb temperature, and then spent vaporiz-
ter understand the process, reduce the number of experiments during ing the solvent, Qout is the heat coming out of the equipment
development of a new product and also in the energy optimization of by means of gas drying, and finally Qloss is the heat lost to the
drying stages. Additionally, a model that extends information beyond exterior. The mentioned heats have the following equations:
the inline/online data is also industrially very appealing because of
time-consuming activities of off-line determinations.
Qin = Fdrying × Cpgas × Tin (2)
Many theoretical frameworks and reviews have been establish for
the use of spray drying (Bhandari et al., 1997; Gharsallaoui et al., 2007;
Kemp et al., 2016) but the next step is to create software’s using all    
theoretical information to help troubleshooting and anticipate process Qfeed = Ffeed 1 − Cfeed × Hvap + Ffeed 1 − Cfeed
scenarios within an industrial environment. Therefore, the objective  
of the present work was to develop a spray drying model that could be ×Cpfeed × Twb − Tfeed (3)
used by any scientist in any project in the food industry and validate
its use. Additionally, the model was further extended to help anticipate
possible challenges during the process itself. After model development Qout = Fdrying × Cpgas × Tout (4)
and implementation, the simulations were verified and validated for
different food products by drying processes with different conditions. Qloss = U × Asup × Tw (5)

2. Materials and methods where, Fdrying is the drying gas flow rate, Cp is the specific
heat, Tin is the drying gas inlet temperature and Tout is the
2.1. Spray drying equipment drying gas outlet temperature, Ffeed is the feed flow rate, Cfeed
the solids concentration of the feed, Hvap is the vaporization
The equipment to be characterized is a Labmaq, model SD10 enthalpy of the solvent in the feed, Twb is the wet bulb tem-
with drying capacity of 10 L of water per hour. The blower has perature and Tfeed the feed temperature. U is the heat transfer
a capacity to pump atmospheric air up to 8 m3 /min (580 kg/h coefficient value across the walls of the equipment, Asup is
Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60 51

Fig. 1 – Schematic representation of spray drying apparatus along with model equations.

the surface area of the equipment and Tw is the logarith- the water intake by the drying gas typically atmospheric air,
mic mean of the temperatures along the spray dryer and the which can either be dehumidified or not, and by means of the
external environment. Substituting the Eqs. (2)–(5) on Eq. (1) feed which is atomized within the dryer. At the outlet we will
and rearranging in relation to the desired output we have the have part as residual solvent in the solid product and another
following equations: part transported as water vapor by the drying gas. Since it is
To determine Tin or Tout : possible to determine or impose a certain outlet temperature
     
Ffeed 1 − Cfeed × Hvap × Ffeed 1 − Cfeed × Cpfeed × Twb − Tfeed + U × Asup × Tw
Tin − Tout = (6)
Fdrying × Cpgas

To determine Fdrying :

     
Ffeed 1 − Cfeed × Hvap × Ffeed 1 − Cfeed × Cpfeed × Twb − Tfeed + U × Asup × Tw
Fdrying = (7)
Cpgas (Tin − Tout )

To determine Ffeed :
value in the energy balance, it is also possible to obtain an
Fdrying × Cpgas × (Tin − Tout ) − U × Asup × Tw
Ffeed =       (8) estimate of the relative humidity at the outlet of the dryer. A
1 − Cfeed × Hvap + Cpfeed × Twb − Tfeed mass balance to the component water will assume Eq. (11).

Specific heats of the drying gas and product to be fed to the


min,gas + min,feed = mout,solid + mout,gas (11)
dryer are a function of the temperature whereby Eqs. (9) and
(10) were used to estimate their values. In the case of the feed
where, m in,gas is the mass of water carried by the drying gas
product an approximation was made by only considering the
at the inlet, min,feed is the mass of water present in the feed,
specific heat of the water since this is the solvent present in
mout,solid is the mass of residual water in the dried solid and
food products.
mout,gas is the mass of water carried by the drying gas at the
  outlet. For simulation purposes, it will be considered that the
Cpgas = 4 × 10−7 × Tin
2
+ 1 × 10−5 × Tin + 1.0049 kJ kg−1 K−1
solid does not have residual water so the expanded version of
(9) Eq. (11) is as follows:

 
−5 −2
 −1 −1
 Fdrying × Hin + Ffeed × 1 − Cfeed = Fdrying × Hout (12)
Cpfeed = 2 × 10 × Tin − 0.0019 × Tin + 4.2258 kJ kg K

(10) where Hin and Hout are the moisture contents of the drying
gas at the inlet and outlet respectively. By determining the
moisture content at the outlet and knowing the value of the
2.2.2. Mass balance drying gas temperature at the outlet determined by Eq. (6), it is
For the spray drying process, the mass balance can take two possible to determine the saturation pressure of water vapor
forms: the mass balance for the dried product and the mass (Pvsat ), the vapor pressure of the drying gas (Pvout ) and finally
balance for the solvent water. It is considered more impor- the relative humidity of the drying gas at the outlet of the
tant to determine the conditions under which the product will equipment (% RHout ), using the following equations:
be dried hence the focus is given on determining the relative
humidity of the drying gas as this is an indicator of the driving Hout × Pchamber
Pvout = (13)
force of the drying. Consequently, it is important to reconcile Hout − 0.62198
52 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

 
77.345+0.0057×Tout − 7235
T
major challenges of spray drying processes since it provides
e out
Pvsat = 8.2
(14) agglomeration and accumulation of product inside the dry-
Tout ing chamber, significantly reducing process throughput, but
also results in products with high hygroscopicity and poor
PVout
%RHout = (15) flowing (Fazaeli et al., 2012). Another important feature of
PVsat
the glass transition temperature is related to the stability of
2.3. Model extension the dried powder during storage since rubbery state materials
have higher molecular mobility resulting in phase separa-
Using the equations proposed in Section 2.2 it is possible to tion of food contents (Ferrari et al., 2013). Since food products
estimate important parameters for the establishment of a presents high content of low molecular weight molecules,
thermodynamic operating space such as the inlet tempera- such as sugars and organic acids, their glass transition tem-
ture, the outlet temperature and the feed rate. The relative peratures tends to be low and the presence of water tends to
humidity data allows to avoid the production of solid parti- lower even further as it acts as a plasticizer. Typically a car-
cles with excessive humidity as these have the tendency to rier agent such as maltodextrin, gum arabic, starches, with
stick to the walls in the dryer. higher glass transition temperature is used to overcome this
sort of challenges. Additionally, an outlet temperature below
2.3.1. Wet bulb temperature and dew-point temperature the glass transition temperature of the product can also be
The determination of these temperatures is important from used. Glass transition temperature can be determined exper-
the operational point of view since these can directly affect the imentally by differential calorimetry, can be estimated using
quality of the product or the process performance. While dry- the Gordon–Taylor equation (Gordon and Taylor, 1952).
ing, the product droplet is heated from its initial temperature,
typically ambient temperature, to equilibrium evaporation x1 Cp1 Tg1 + x2 Cp2 Tg2
Tgfeed = (18)
temperature. During this period, removal of moisture from the x1 Cp1 + x2 Cp2
surface of the droplet follows a constant rate thereby main-
where x1 and x2 are the solute and water mass fractions
taining the drop at a constant temperature due to the latent
present in the feed, Tg1 and Tg2 are the glass transition tem-
heat of vaporization. Since the surface of the droplet is sat-
peratures of pure product and water, and Cp is the specific
urated with moisture, the maximum temperature reached is
heat change during the glass transition (Loerting et al., 2015). It
the wet bulb temperature (Twb ), and when the constant rate
is possible to find some Tg values for food materials in another
period is over, the powdered solid product will reach the outlet
work (Tontul and Topuz, 2017) whereas the values of the spe-
temperature of the drying gas. Since the outlet temperature
cific heat variation can be estimated according to the molar
is typically only about 12% above the wet bulb temperature,
mass of the sugars (Avaltroni et al., 2004). It has been demon-
the product never reaches the high values of the inlet tem-
strated elsewhere that to avoid “stickiness” problems within
peratures. Therefore, the determination of this temperature
the drying chamber, the temperature of the drying gas at the
is important for the scientist in order to avoid degradation
outlet should be about 10–20 ◦ C below the estimated value of
or alteration of the organoleptic characteristics by the action
Tg (Woo et al., 2008).
of temperature (Anandharamakrishnan et al., 2007). The wet
bulb temperature can be estimated by Eq. (16):
2.3.3. Estimation of droplet diameter and final particle size
(Hout − Hin ) × Hvap Atomization is fundamental on spray-drying drying process
Twb = Tin − (16) since it’s the step where the product is divided into droplets
Cpgas
small enough to dry during the residence time inside the dryer.
On the other hand, the dew point temperature (Tdp ) Therefore, in addition to defining the final size of the powder, it
is of extreme importance when considering the industrial also has a determinant role in drying kinetics, since the expo-
operation since many problems can come from clogging or sure area for mass and heat transfer will be determinant for
excessive moisture at the collection point of the dry prod- effective drying. There are three main types of atomizers that
uct (Gianfrancesco et al., 2008). Thus, if the collection point can be found in spray-dryers: rotary, pressure nozzles and two-
is exposed to external temperatures low enough to lower the fluid nozzles. A comparison of these types of atomizers can be
drying gas outlet temperature to the dew-point, then what found in other works (Walzel, 2011). The drop formation is a
happens is that the water carried by the gas will condensate combination of factors such as the energy applied, the noz-
and wet the product, causing batch loss or clogging of the zle geometry and the feed flow rate. Thus, separation of the
cyclone, or the bag filter. The dew point temperature can be liquid into drops involves a balance between disruptive forces
determined using the Magnus equation (Eq. (17)), where con- and cohesive forces of viscosity and surface tension. For each
stants b and c have values of 17.62 and 243.12 ◦ C respectively. of these types of atomizers there are numerous equations to
estimate the droplet size, being here some of the simplest to
 %RHout  b×Tout
c. ln + use. For rotative nozzles (Masters, 1979):
Tdp =  %RH
100
 c+Tout
(17)
b − ln out
100 + b×T out
c+Tout 0.15
D50 = 0.008 × Ffeed × D−0.18 × N−0.05 × ω−0.75 × 0.07
feed
(19)
2.3.2. Estimation of the glass transition temperature
The glass transition temperature is a second order thermal and, for pressure nozzles (Lefebvre, 1988):
phenomenon where the matrix of a material transits from a  −0.25
soft and rubbery state at higher temperatures to a rigid and 0.25
D50 = 4.0 × Ffeed × P−0.5
feed
× feed × feed 0.07
× feed (20)
vitreous state at lower temperatures and can be associated
with material stickiness, wall deposition and consequently where D is the disk diameter, N is the number of blade s, ω the
low yield (Adhikari et al., 2005). This “stickiness” is one of the blade speed, feed ,  feed , Pfeed and feed are the viscosity, sur-
Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60 53

face tension, pressure and density of the liquid product being 2.3.5. Drying kinetics
fed. The type of nozzle available during the validation experi- To establish the equation for the water evaporation rate from
ments of the present is the two fluid nozzle. Using these type of the product droplet some considerations are needed. Firstly,
nozzles, the formation of the droplet is governed by the feed it is necessary to consider that the relative velocity between
flow rate, the atomizing gas flow rate and the nozzle geom- the droplet and the drying gas can be neglected and conse-
etry, which in combination establish a relationship between quently the evaporation mechanism occurs in stagnant gas
the relative velocities of the feed product and the atomizing so the Nusselt number (Nu) is equal to 2. Secondly, only the
gas. Once again there are numerous equations to carry out the period of constant evaporation rate is considered. During this
proposed estimation, but some caution is required in because period the rate of the droplet shrinkage can be equalized to
there are different nozzle geometries, and some adjustments the mass flow out of the droplet (Schiffter and Lee, 2007):
in the empirical parameters are usually is required. Thus the
equation proposed for the pneumatic nozzle in the present dVdroplet h × (Tout − Twb )
droplet = −Adroplet (25)
work is (Lubanska, 1970): dt Hvap

  0.5 considering a spherical droplet and Nu = 2, then


feed Ffeed
D50 = Kd Dn 1+ (21)
gas × We Fatom dDdroplet 2 × kgas × (Tout − Twb )
Ddroplet = (26)
dt droplet Hvap

v2gas × feed × Dn
We = (22) integrating Eq. (25) it is possible to determine the time required
feed to reduce the initial droplet size to the final particle size.

where Kd is a constant, Dn is the orifice diameter in the atom- droplet × Hvap  


t= D2droplet − D2particle (27)
izer, feed and gas are the kinematic viscosities of the product 2 × kgas × (Tout − Twb )
and gas (m2 /s), Ffeed and Fgas are the feed and drying gas flow
rates (kg/s). We is the Weber number relating fluid’s inertia 2.4. Model implementation
with its surface tension, where vatom is the velocity of the
atomizing gas at the exit of the atomizing nozzle and  is the The model was implemented using a spreadsheet software.
surface tension of the feed product. These equations estimate The selection of this type of software is related to the sim-
the value of D50, which is the diameter that divides half of the plicity of use, not requiring training in code development, and
droplet population, i.e., half has a diameter lower then D50 and widespread use in industry.
the other half a diameter above the estimated value. Addition-
2.5. Spray drying characterization
ally, the final particle size (Dp) can be given by the following
expression (Vehring, 2008):
To determine the overall coefficient of heat transfer, experi-
 ments with no drying were performed, i.e. no feed product
Cfeed feed was dried throughout these experiments. Therefore the dif-
Dparticle ∼
3
= Ddroplet (23)
Csolid solid ference between inlet temperature and outlet temperature of
the drying gas will only be caused by energy loss (Qloss ). The
This expression relates the estimated droplet size with the set-points imposed on the equipment were the drying gas flow
final solid size, which is smaller since the product is being rate which varied between 2, 3, 5 and 8 m3 /min while the inlet
dried. temperature of the drying gas was another set point varying
between 50, 100 and 150 ◦ C. With the results, an empirical
model was established allowing the determination of the over-
2.3.4. Droplet size distribution
all coefficient of heat transfer accordingly to the conditions
The heterogeneous nature of the atomization process caused
used.
by the different mechanisms of separation and union of the
atomized liquid results in droplets that vary in size. Therefore
2.6. Model validation
spray nozzles do not produce uniform sprays, with droplets of
identical size in any condition of use. Instead, a spray should
In order to validate the proposed model, experiments were
be viewed as a spectrum of arbitrarily distributed droplet sizes
conducted where different food products were dried using an
around an average value determined in the previous section
industrial spray dryer together with a cyclone in open cycle
(Lefebvre, 1989). There is still no theory that allows determin-
mode since atmospheric air was used as drying gas. The food
ing exactly the size distribution of droplets, using probabilistic
products dried were bovine milk, goat milk, coffee and acerola
functions that allow the mathematical representation of the
fruit juice, with initial concentrations of solids ranging from 4
distribution. Although there are several distributions, one of
to 15% (w/w). Feed flow rates ranged from 1 kg/h to 5 kg/h and
the most used is Rosin-Rammler (Eq. (24)) (Grant et al., 1993):
drying gas flow rates ranged from 144 kg/h to 580 kg/h. The
inlet temperatures used ranged from 50 ◦ C to 150 ◦ C and the
 −x n
outlet from 30 to 90 ◦ C. For the atomization, a two fluid nozzle
Y =1−e x0
(24)
with an internal diameter of 2 mm was used, with atomization
gas flow rates ranging from 2 to 4 kg/h. Table 1 summarizes
where Y is the cumulative fraction of material by weight the information of the experiments performed for validation.
smaller than size x, n is the uniformity constant and x0 is While spray drying, the equipment was initially stabilized only
the characteristic particle size defined as the size 63,2% of with the passage of drying gas until it reached the set point
particles are smaller. of inlet temperature or the set point of outlet temperature.
54 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

Table 1 – Comparison between the estimated and experimental values for Tin , Tout and RHout.
Trial Fdrying (kg/h) Ffeed (kg/h) Est. Tin (◦ C) Exp. Tin (◦ C) Est. Tout (◦ C) Exp. Tout (◦ C) Est. RHout (◦ C) Exp. RHout (◦ C)

1 580 ± 6 1.0 109 111 ± 0.37 90 90 ± 0.33 4 3 ± 0.2


2 580 ± 6 3.0 119 120 ± 0.39 90 90 ± 0.33 5 5 ± 0.14
3 580 ± 6 5.0 127 128 ± 0.41 90 90 ± 0.33 5 6 ± 0.26
4 580 ± 6 1.0 145 146 ± 0.43 120 120 ± 0.39 1 1 ± 0.08
5 580 ± 6 3.0 154 155 ± 0.45 120 120 ± 0.39 2 3 ± 0.15
6 580 ± 6 5.0 163 163 ± 0.47 120 120 ± 0.39 2 3 ± 0.23
7 580 ± 6 2.0 120 120 ± 0.39 84 83 ± 0.31 4 5 ± 0.08
8 580 ± 6 2.0 170 170 ± 0.49 105 103 ± 0.35 2 2 ± 0.133
9 580 ± 6 2.0 120 120 ± 0.39 83 83 ± 0.31 4 4 ± 0.08
10 580 ± 6 2.0 170 170 ± 0.49 105 103 ± 0.35 2 3 ± 0.14
11 580 ± 6 2.0 145 145 ± 0.44 94 92 ± 0.34 4 4 ± 0.12
12 144 ± 2 1.2 102 100 ± 0.35 50 50 ± 0.25 27 27 ± 0.15
13 288 ± 3 1.2 73 70 ± 0.29 50 50 ± 0.25 26 26 ± 0.23
14 144 ± 2 1.2 102 100 ± 0.35 50 50 ± 0.25 31 27 ± 0.48
15 288 ± 3 1.2 73 70 ± 0.29 50 50 ± 0.25 26 26 ± 0.26
16 288 ± 3 1.2 120 120 ± 0.39 84 83 ± 0.31 7 9 ± 0.37
17 288 ± 3 1.2 110 110 ± 0.37 76 76 ± 0.3 9 10 ± 0.13
18 288 ± 3 1.2 72 72 ± 0.29 48 50 ± 0.25 28 26 ± 0.33
19 288 ± 3 1.2 80 80 ± 0.31 54 55 ± 0.26 20 21 ± 0.38
20 288 ± 3 1,2 100 100 ± 0.35 74 74 ± 0.3 10 11 ± 0.21

The equipment was considered to be stabilized or at steady- 3. Results and discussion


state when the temperature did not varied by more than 1 ◦ C
for 10 min. After this stabilization, the addition of pure water 3.1. Global heat transfer coefficient determination
to the equipment was started, at a given flow rate and when
a new steady state was reached the pure water supply was The overall heat transfer coefficient of the equipment, U is a
switched to product. The experimental values were recorded measure that expresses the conductive and convective bar-
5 times throughout the experiments. riers for the transfer of heat through the dryer walls. The
determination of this coefficient is decisive for the accuracy of
2.7. Drying gas flow rate and temperature the model and assumes different values for different equip-
measurements ments. To determine U, the spray dryer was used only with
the passage of the drying gas and the difference between the
The spray drying equipment has a volumetric flow rate gas inlet and outlet temperature of the equipment was measured
meter, with and accuracy of 1%, prior to the resistance when the operation reached a steady state, avoiding tempera-
whereby the measured values are the volume of gas passing ture variations in the wall of the equipment. The difference
through at room temperature. The conversion of the volume to between the inlet and outlet temperature, in these experi-
mass is done using the atmospheric air density at room tem- ments is only caused by the loss of heat to the exterior, being
perature. The measurement of the inlet temperature is made relatively easy to determine U knowing the surface area of the
shortly after the exit of the electrical resistance of the equip- equipment. From Eq. (1), Qfeed equals zero, then:
ment and immediately before the drying gas distributor, while
the outlet temperature of the drying gas is made just before the Qloss = Qin − Qout (26)
cyclone. Both temperature sensors are resistive type, PT100
class A.
expanding Eq. (26), U can be determined by Eq. (27):

2.8. Relative humidity measurements Fdrying × Cpgas × (Tin − Tout )


U= (27)
Asup × Tw
To measure the relative humidity, a portable digital hygrome-
ter, model MTH-1362, Brand Ininipa, with a 2.5% accuracy was
U values for the industrial spray dryer studied varied
used. Measurements were taken at the exit of the drying gas
between 5.0 and 2.2 W/m2 K. These results are below to those
at the top of the cyclone. The sensor has a relative humid-
reported by Saravacos and Maroulis (2001) that reported val-
ity range ranging from 0 to 100% and a K-type thermocouple
ues for drying with air between 20–100 W/m2 K not specifying
with a measuring range of −50 ◦ C to 1000 ◦ C was used to avoid
the type of equipment used (Saravacos and Maroulis, 2001).
limitations on the outlet temperature of the drying gas.
U values increase with increasing flow rates of the drying gas
and when the outlet temperatures are high (above >120 ◦ C), but
2.9. Particle size determination remain relatively stable at lower temperatures. These results
suggest greater dependence on the temperature than on the
Dry powder samples from experiments were characterized drying gas flow rate. Eq. (28) is a quadratic regression used to
using scanning electronic microscopy, Tescan Vega 3, at 15 kV estimate the U value used in all simulations henceforward.
without metal plating. The micrographs were analyzed using
imageJ software to determine the diameters of each visible
U = 8.65 − 0.009Fdrying − 0.057Tin + 6.14 × 10−6 Fdrying
2
powder particle. Measurements were preformed in triplicate
for all samples. +4.2 × 10−5 Fdrying Tin − 0.0001Tin
2
(28)
Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60 55

the experimental values for different temperature ranges,


between 170 ◦ C and 45 ◦ C. The linear equations fitted to the
data revealed slopes close to 1, (Experimental Inlet Tem-
perature = 1.0255 * Estimated Inlet Temperature − 2.66, Experi-
mental Outlet Temperature = 0.981 * Estimated Outlet Temper-
ature + 1.245) which indicates the proximity of the estimated
values with the experimental values. For both parameters,
all points are very close to the 95% confidence limit around
the compared values. The biggest difference between the esti-
mated and the experimental values for inlet temperature were
recorded at trial 13 and 15 with a difference of 3 ◦ C while for
outlet temperature, trial 10 and 11 recorded a difference of 2 ◦ C
and −2 ◦ C for trial 18. The mean absolute differences between
the estimated values and the experimental values varied from
0.8 ◦ C for the inlet temperature to 0.55 ◦ C for the outlet temper-
ature. These small variations may be related to the estimates
of U but also to the approximation used for specific heat and
enthalpy of water despite the use of food products. Heat bal-
ance was only validated for one equipment, yet it should be
robust enough to be applied in other equipment since the
value of U for that equipment is used. The fact that different
values of drying gas flow rate and feed flow rate were used,
supports the robustness and the versatility of the model.

3.3. Model validation: mass balance

Spray drying mass balance was modeled with the following


approximation: the drying gas removed 99% of the water and
thus the dry product would only have 1% of water content. This
approximation allowed the estimation of the outlet relative
humidity value as this variable is of paramount importance
for the drying kinetics for two reasons. Firstly, drying gas car-
rying more water decreases the mass transfer driving force
out of the droplet and thus increases the time required to dry.
Secondly because it lowers the glass transition temperature
of amorphous sugars and since the moisture acts as a plasti-
cizer, stickiness and wall deposition can occur or the shelf life
Fig. 2 – Representation of the difference between the
of food products maybe reduced (Mensink et al., 2017). Low
experimental values and the estimated values for the inlet
outlet relative humidity (%RHout ) can be achieved by simply
(left) and outlet (right) temperatures.
increase the outlet temperature. Such a solution may be inef-
fective because it consumes more energy and can thermally
degrade products and/or change their sensory characteristics.
Therefore, in order to define the drying conditions of a given
3.2. Model validation: heat balance product, the scientist should make a balance between the
temperatures at the outlet, the relative humidity and a satis-
A simple model was developed as described in Section 2 and factory throughput. Typically the spray drying equipment does
was completed with the determination of the global heat not have instrumentation to measure the relative humidity
transfer coefficient. Consequently the model needs to be ver- so it is not integrated at any control system increasing the
ified and validated. First, the verification of a model involves importance of simulating these values through models. For
the critical analysis of the values obtained by the simulations, validation, relative humidity measurements were made at the
that is, if the values obtained are in agreement with what exit of the cyclone with a digital hygrometer using the drying
is expected. Such a check first allows to filter possible gross conditions described in Table 1 and the values were compared
errors of model development or calculations (Robinson, 2014). with the simulations. Fig. 3 shows the comparison between
Secondly, the validation of the model involves the execution the estimated relative humidity values and those observed
of experimental drying tests where the output of the model is experimentally.
compared with experimental results. Fig. 2 shows the compar- Values of outlet relative humidity were obtained in the
ison between experimental measurements, Tin and Tout , and range of 8%–27%, which covers a good part of the typical
the output of the model simulations. operating space of food processing (Mahdavi et al., 2014). Com-
The experiments had inlet temperatures ranging from parison between the estimated RH out with the experimental
70 ◦ C to 160 ◦ C, while the outlet temperatures ranged from measurements by fitting a linear equation to the data, (Exper-
50 ◦ C to 105 ◦ C. This temperature range is in accordance imental RH out = 0,932 * Estimated RH out + 0.89) returned a
with the typical operating’s spaces used in food processing slope of 0.932 which is close to 1 indicating the proximity of
(Gharsallaoui et al., 2007; Shishir and Chen, 2017). Inlet and both values. Most of the data points are inside the 95% confi-
outlet temperatures estimated by the model correspond to dence interval. The biggest difference recorded was −4.0% for
56 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

Fig. 3 – Comparison between the estimated relative Fig. 5 – Simulation of the drying effect on the glass
humidity with the experimental values. transition temperature for different solutes used in food
industry.

The contour plot was generated using a polynomial


equation to fit data (Eq. (29)) presenting a coefficient of deter-
mination of 96.75. A geometric figure was drawn inside the
area to present the thermodynamic design space. The bound-
aries of the geometric figure are also boundaries for a spray
drying process or undesirable operational areas and are sum-
marized in Table 2:
 
Ffeed Ffeed
%RH = 81.5 + 1485.4 · − 13, 264 ·
Fdrying Fdrying

Ffeed
2
− 1.5 · Tout + 0.007 · (Tout )2 − 10.5 · Tout · (29)
Fdrying

Fig. 4 – Contour plot of the combined design space used in


all trials.
3.5. Glass transition estimative
trial 14, but the mean of absolute differences is only 0.85%.
Such a difference may be related to the approach made by the Glass transition temperature estimations are relevant while
model that considers the solid product to be totally dry, which planning experiments since it has been shown that this tem-
in fact does not happen, being part of the water retained in perature is related to the stickiness of particles to the dryer
the solid rather than being carried by the drying gas. Tonon wall affecting the process yield, which industrially can be crit-
et al. (2008) reported higher percentages of residual moisture ical. Fig. 4 shows the use of Eq. (18) used to estimate the value
on spray dryed Açai while using a combination of parame- of the glass transition temperature for a mixture of solutes
ters (Trial 3, Tin = 150 ◦ C and Ffeed = 21 g/min) that yielded the with water. For Fig. 4, the solutes, fructose, maltodextrin and
highest RHout estimated by our model, supporting the relation starch were isolated and dissolved in water and the mass frac-
between RHout and residual moisture. tion of each solute was varied, simulating the effect that the
drying would have on the glass transition temperature (Fig. 5).
3.4. Design space From the simulation, it is verified that the glass transition
temperature of the chosen solutes varies from values close to
The combination of mass and heat balance allows the map- −80 ◦ C to values close to 0 ◦ C for fructose, 150 ◦ C for maltodex-
ping of a thermodynamic operating space for a given process, trin DE10 and above 200 ◦ C for starch. The most relevant glass
which is important to support decisions on drying parameters transition temperature are those obtained for mass fractions
since they critically affect properties of the final solid prod- close to 1%wt since the final product has only a small mass
uct (Lebrun et al., 2012). The inlet and outlet temperature, the fraction of water. Results can be verified by the estimates of
ratio between feed flow rate and drying gas flow rate and the the glass transition temperature. Since water acts as a plas-
relative humidity of drying gas at the outlet can be used to ticizer, the expected result is an increase on the transition
define the design spaces of all trials performed in the present temperature with the decrease of the water mass fraction as
work (Fig. 4). Such representation can also help the scale-up demonstrated by Fig. 4. Additionally, the selected solutes have
of a process using non-dimensional parameters such as RHout significant variations of molar mass which generates increas-
and Ffeed /Fdrying ratio. ing results of glass transition temperature. Although no data
Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60 57

Table 2 – Description of spray drying process boundaries.


Boundary Description

A Low throughput area where Ffeed is too low and thus process provides low benefit and is inefficient
B Outlet temperature is too high which may cause product thermal degradation;
C Equipment limitations since inlet temperature required is too high;
D Dew point temperature is to high and condensation may occur;
E High relative humidity may cause insufficient drying and product losses

is presented to validate the equation proposed in the present


study, this is in accordance with data in the literature, which
allows us to verify that the estimates are correct (Adhikari
et al., 2004, 2005; Normand et al., 2013).

3.6. Droplet size estimation and particle size


distribution

The atomization nozzle available in the spray drying used for


the present work is a two fluid nozzle and therefore the equa-
tion described in the respective section of this nozzle was used
(Eq. (23)). This equation has a constant K that can be used as
correction factor to better fit the data. The equation allows to
estimate the droplet size but as the droplet is dried it’s size is
reduced becoming difficult to validate this estimate without
proper in-line characterization tools. Thus and maintaining
the simplified philosophy of the present work Eq. (21) was
used using the estimated droplet size, and the final particle
size was estimated. Fig. 6 shows the comparison between the
estimated values for the particle size and the values obtained
experimentally.
The experimentally obtained particle sizes varied between
11 ␮m and 45 ␮m and these values were within the range of
atomization nozzles (Vicente et al., 2013). The estimated val-
ues for the particle size are very close to the experimental
values of particle size. Fitting a linear equation (Experimen-
tal Particle Size = 0.87 * Estimated Particle Size + 3.84), a slope
of 0.87 was obtained, revealing the similarity between both
values. The biggest difference recorded was 5.4 ␮m, but the
mean of absolute differences was 2.2 ␮m which supports the
accuracy of the equation used. It is important to mention that
drying conditions play a part on the particle size, since it can
affect the final geometry of the particle. Shriveled or ballooned
particles, are dependent of the drying kinetics, and conse-
quently the volume and density of such particles is affected. To
elaborate, if high drying kinetics, as a result from high temper-
atures or low relative humidity, is used then droplets tend to
dry faster, creating a thick shell at the droplet boundary that is Fig. 6 – Representation of the comparison between the
sustained throughtout the process by the internal pressure of experimental particle size with the estimated particle size
the liquid being evaporated. As a result, a large ballooned with (a) and cumulative frequency of the estimated particle size
high volume and low density particle is created. In contrast, using Rosin-Rammler distribution (b).
if low drying kinetics is used as a result of low temperatures
or high relative humidity then a thin layer formed in the ini-
tial stages of drying will tend to follow the retracted surface
until its thickness is stable enough to sustain the structure different particle sizes where the only difference in the dry-
of the particles. As a result a shriveled particle with lower ing process is the outlet temperature. Alamilla-Beltran et al.
volume and higher density is created. The evidence of the (2005) also report similar results. A more detailed explanation
mentioned shell can be found in Elversson work (Elversson can be found elsewhere (Vehring, 2008; Vicente et al., 2013).
et al., 2003). Peclet number is a dimensionless number that This explains why the same atomization conditions results in
relates the advective transport rate with the diffusive trans- different particle sizes. Particle size distribution can be esti-
port rate and can be used to relate the drying conditions, mated using a log-normal equation (Eq. (24)) with mean values
since in high drying kinetics the advective transport rate is of particle size as demonstrated by Fig. 6b, representing trial 3.
higher then the diffusive transport rate whereas in low drying The estimated curves show a good fit with other particle size
kinetics the contrary occurs. Vicente et al. (2013) clearly report values such as D10 and D90.
58 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

Table 3 – Model estimates for spray drying different food products.


Trial Food product Fdrying (kg/h) Ffeed (kg/h) Est. Tin (◦ C) Est. Tout (◦ C) Est. Twb (◦ C) Est. Tdp (◦ C) Est. RHout (%) Est. tdrying (ms)

1 Goat milk 580 1.0 111 90 86 25 4 65


2 Goat milk 580 3.0 120 90 79 27 5 100
3 Goat milk 580 5.0 128 90 73 29 5 130
4 Goat milk 580 1.0 146 120 116 26 1 40
5 Goat milk 580 3.0 155 120 109 28 2 70
6 Goat milk 580 5.0 163 120 103 30 2 100
7 Cow milk 580 4.0 106 83 76 26 4 120
8 Cow milk 580 4.0 132 103 96 26 2 100
9 Cow milk 580 4.0 106 83 76 26 4 850
10 Cow milk 580 4.0 132 103 96 26 2 690
11 Cow milk 580 4.0 119 92 85 26 4 250
12 Coffee 144 1.2 100 50 32 28 27 90
13 Coffee 288 1.2 70 50 41 25 26 110
14 Coffee 144 1.2 100 50 33 28 31 20
15 Coffee 288 1.2 70 50 41 25 26 20
16 Acerola juice 288 1.2 120 83 74 26 7 70
17 Acerola juice 288 1.2 110 76 67 25 9 70
18 Acerola juice 288 1.2 72 50 41 25 28 110
19 Acerola juice 288 1.2 80 55 46 25 20 100
20 Acerola juice 288 1.2 100 74 59 25 10 80

ing outlet temperature decreases drying time due to higher


heat transfer to the droplet. Zbicinski et al. reported a series
of extensive experiments on spray drying kinetics and particle
residence time using in situ advanced testing equipment. They
also reported drying times shorter than 1 s and concluded that
drying gas temperature and atomization conditions have a
profound effect on drying kinetics just as the presented data
suggests (Zbicinski et al., 2002).

3.8. Further verification and validation

The proposed model, although simple, allows the estimation


of many parameters that aid in the decision making process
of spray drying. Table 3 summarizes most values estimated
by the proposed equations in Section 2. Not all data pre-
sented can be validated, but only verified. Thus the values
obtained are in agreement with the results of several works
already published in the food area. Reddy et al., spray dried
goat milk using inlet/outlet air temperatures in the range
of 160-170–180 ◦ C/85 ◦ C in combination with increasing solids
Fig. 7 – Contour plot of the influence of outlet temperature
concentration and evaluated the influence on the physical
and droplet size on the minimum time to dry a droplet.
properties of the dried milk. The authors report a constant
outlet temperature and a constant feed flow rate with increas-
3.7. Drying kinetics
ing inlet temperature which is thermodynamically impossible
unless the not mention drying gas flow rate is also decreasing
Under certain considerations it was possible to estimate the
(Reddy et al., 2014). Fonseca et al., also dried goat milk using
drying time of different trials using Eq. (27). As mentioned
similar conditions presented in trial 1 to 6, with atomization
before, the considerations taken were two: drying occurred in
parameters of 1 L/h of feed flow rate with 40–60 L/min of atom-
stagnant gas, with nusselt number equal to 2, and estimated
ization air flow rate. Using Eq. (21) and (22) it is possible to
time corresponds to the constant rate period. The minimum
determine a mean particle size of 13.3 ␮m, a value close to the
time required to completely dry a droplet under the different
D50 value reported by the authors (Fonseca et al., 2011). Fras-
drying and atomization conditions proposed on trials 1 to 20
careli et al. spray dryed coffee emulsion with gum Arabic using
was calculated and then a polynomial equation (Eq. (30)) was
a central composite rotable design variying parameters such
used to fit the data and a contour plot was made (Fig. 7).
as Cfeed (10–30%), Tin (150–190) and oil concentration which is
not used in our model. The authors report values for the out-
tdrying = −15.29 − 0.56 · Tout + 0.006 · (Tout )2
let temperature, that combined with the drying gas flow rate
+2.80 · D50 + 0.02 · (D50)2 − 0.03 · Tout · D50 (30) and the feed flow rate can validate our model for a U value
of 17 J/m2 K s (Superficial Area = 0.235 m2 ), which is a typical
Eq. (30) has a coefficient of determination of 98.69 and all value for a lab scale equipment and is much higher than our
parameters were significative for a p < 0.001. As expected dry- double wall industrial equipment (Frascareli et al., 2012). Park
ing time increases with increasing droplet size due to bigger et al. reported spray drying bovine milk with a constant outlet
volume and less surface area for mass transfer while increas- temperature of 90 ◦ C while inlet temperatures increased var-
Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60 59

ied from 160, 210 and 260 ◦ C. The outlet temperature was kept physical properties of black mulberry juice powder. Food
constant due to increase on feed flow rate. The authors report, Bioprod. Process. 90 (4), 667–675.
even though consider it counterintuitive, that increasing inlet Ferrari, C.C., Marconi Germer, S.P., Alvim, I.D., de Aguirre, J.M.,
2013. Storage stability of spray-dried blackberry powder
temperature benefits flavor but mostly because of solids con-
produced with maltodextrin or gum arabic. Dry. Technol. 31
centration and not higher wet-bulb temperature (Park et al., (4), 470–478.
2016). Moreira et al. spray dried acerola pomace using a Mini Fletcher, D.F., Guo, B., Harvie, D.J.E., Langrish, T.A.G., Nijdam, J.J.,
Spray Dryer Buchi B-290. The authors are not very accurate Williams, J., 2006. What is important in the simulation of
with drying conditions used, reporting 5.51 × 104 kg/h for aspi- spray dryer performance and how do current CFD models
rator flow rate, while the max air aspirator flow rate for this perform? Appl. Math. Modell. 30 (11), 1281–1292,
equipment is approximately 40 kg/h if the drying gas is not http://dx.doi.org/10.1016/j.apm.2006.03.006.
Fonseca, C.R., Bento, M.S.G., Quintero, E.S.M., Gabas, A.L.,
re-used, and since the feed rate is given, the peristaltic pump
Oliveira, C.A.F., 2011. Physical properties of goat milk powder
rate it is an unknown parameter. The inlet temperatures are with soy lecithin added before spray drying. Int. J. Food Sci.
much lower than the ones used in our work but this is proba- Technol. 46 (3), 608–611.
bly because our drying gas flow rate is much higher then the Frascareli, E.C., Silva, V.M., Tonon, R.V., Hubinger, M.D., 2012.
one used with the Buchi B-290 (Moreira et al., 2009). Effect of process conditions on the microencapsulation of
coffee oil by spray drying. Food Bioprod. Process. 90 (3),
413–424.
4. Conclusions Gharsallaoui, A., Roudaut, G., Chambin, O., Voilley, A., Saurel, R.,
2007. Applications of spray-drying in microencapsulation of
The present work developed a simplified model of spray dry- food ingredients: an overview. Food Res. Int. 40 (9), 1107–1121,
ing based on fundamental equations focusing application in http://dx.doi.org/10.1016/j.foodres.2007.07.004.
Gianfrancesco, A., Turchiuli, C., Dumoulin, E., 2008. Powder
food industry. The development of this model allows the
agglomeration during the spray-drying process:
estimation of key parameters for the design of processes measurements of air properties. Dairy Sci. Technol. 88 (1),
avoiding waste with successive laboratory tests throughout 53–64, http://dx.doi.org/10.1051/dst:2007008.
the development of new products by spray drying. The main Gordon, M., Taylor, J.S., 1952. Ideal copolymers and the
features of the proposed model are its simplicity, that it can second-order transitions of synthetic rubbers. I.
be easily applied in spreadsheet software or other free code Non-crystalline copolymers. J. Chem. Technol. Biotechnol. 2
(9), 493–500.
software and that either a trained scientist or a junior scien-
Grant, P., Cantor, B., Katgerman, L., 1993. Modelling of droplet
tist can easily implement it. Model simulations were verified
dynamic and thermal histories during spray forming—II.
and validated presenting precision while estimating process Effect of process parameters. Acta Metall. Mater. 41 (11),
parameters and drying conditions for a large range of food 3109–3118.
products, temperatures and relative humidity. Therefore it is Grasmeijer, N., de Waard, H., Hinrichs, W.L., Frijlink, H.W., 2013. A
concluded that the developed model is precise and robust for user-friendly model for spray drying to aid pharmaceutical
industrial scale equipment in food industry. product development. PLoS One 8 (9), e74403.
Ivey, J.W., Vehring, R., 2010. The use of modeling in spray drying
of emulsions and suspensions accelerates formulation and
References process development. Comput. Chem. Eng. 34 (7), 1036–1040.
Kemp, I.C., Hartwig, T., Herdman, R., Hamilton, P., Bisten, A.,
Bermingham, S., 2016. Spray drying with a two-fluid nozzle to
Adhikari, B., Howes, T., Bhandari, B., Troung, V., 2004. Effect of
produce fine particles: atomization, scale-up, and modeling.
addition of maltodextrin on drying kinetics and stickiness of
Dry. Technol. 34 (10), 1243–1252.
sugar and acid-rich foods during convective drying:
Lebrun, P., Krier, F., Mantanus, J., Grohganz, H., Yang, M., Rozet, E.,
experiments and modelling. J. Food Eng. 62 (1), 53–68.
Boulanger, B., Evrard, B., Rantanen, J., Hubert, P., 2012. Design
Adhikari, B., Howes, T., Lecomte, D., Bhandari, B., 2005. A glass
space approach in the optimization of the spray-drying
transition temperature approach for the prediction of the
process. Eur. J. Pharm. Biopharm. 80 (1), 226–234.
surface stickiness of a drying droplet during spray drying.
Lefebvre, A., 1988. Atomization and Sprays, vol. 1040. CRC press.
Powder Technol. 149 (2), 168–179.
Lefebvre, A.H., 1989. Properties of sprays. Part. Part. Syst. Char. 6
Alamilla-Beltran, L., Chanona-Perez, J.J., Jimenez-Aparicio, A.R.,
(1–4), 176–186.
Gutierrez-Lopez, G.F., 2005. Description of morphological
Loerting, T., Fuentes-Landete, V., Handle, P.H., Seidl, M.,
changes of particles along spray drying. J. Food Eng. 67 (1),
Amann-Winkel, K., Gainaru, C., Böhmer, R., 2015. The glass
179–184.
transition in high-density amorphous ice. J. Non-Cryst. Solids
Anandharamakrishnan, C., Rielly, C., Stapley, A., 2007. Effects of
407, 423–430.
process variables on the denaturation of whey proteins during
Lubanska, H., 1970. Correlation of spray ring data for gas
spray drying. Dry. Technol. 25 (5), 799–807.
atomization of liquid metals. JOM 22 (2), 45–49.
Avaltroni, F., Bouquerand, P., Normand, V., 2004. Maltodextrin
Mahdavi, S.A., Jafari, S.M., Ghorbani, M., Assadpoor, E., 2014.
molecular weight distribution influence on the glass
Spray-drying microencapsulation of anthocyanins by natural
transition temperature and viscosity in aqueous solutions.
biopolymers: a review. Dry. Technol. 32 (5), 509–518.
Carbohydr. Polym. 58 (3), 323–334.
Maltesen, M.J., Bjerregaard, S., Hovgaard, L., Havelund, S., van de
Baldinger, A., Clerdent, L., Rantanen, J., Yang, M., Grohganz, H.,
Weert, M., 2008. Quality by design — spray drying of insulin
2012. Quality by design approach in the optimization of the
intended for inhalation. Eur. J. Pharm. Biopharm. 70 (3),
spray-drying process. Pharm. Dev. Technol. 17 (4), 389–397.
828–838, http://dx.doi.org/10.1016/j.ejpb.2008.07.015.
Bhandari, B.R., Datta, N., Howes, T., 1997. Problems associated
Masters, K., 1979. Spray Drying Handbook. George Godwin
with spray drying of sugar-rich foods. Dry. Technol. 15 (2),
Limited, John Wiley and Sons, London, New York.
671–684.
Mensink, Maarten A., Frijlink, Henderik W., van der Voort
Elversson, J., Millqvist-Fureby, A., Alderborn, G., Elofsson, U., 2003.
Maarschalk, Kees, Hinrichs, Wouter L.J., 2017. How sugars
Droplet and particle size relationship and shell thickness of
protect proteins in the solid state and during drying (review):
inhalable lactose particles during spray drying. J. Pharm. Sci.
mechanisms of stabilization in relation to stress conditions.
92 (4), 900–910.
Eur. J. Pharm. Biopharm. 114, 288–295, ISSN 0939-6411
Fazaeli, M., Emam-Djomeh, Z., Ashtari, A.K., Omid, M., 2012.
https://doi.org/10.1016/j.ejpb.2017.01.024.
Effect of spray drying conditions and feed composition on the
60 Food and Bioproducts Processing 1 0 7 ( 2 0 1 8 ) 49–60

Moreira, G.E.G., Costa, M.G.M., de Souza, A.C.R., de Brito, E.S., de Shishir, Mohammad Rezaul Islam, Chen, Wei, 2017. Trends of
Medeiros, M.d.F.D., de Azeredo, H.M.C., 2009. Physical spray drying: a critical review on drying of fruit and vegetable
properties of spray dried acerola pomace extract as affected juices. Trends Food Sci. Technol. 65, 49–67, ISSN 0924-2244
by temperature and drying aids. LWT—Food Sci. Technol. 42 https://doi.org/10.1016/j.tifs.2017.05.006.
(2), 641–645. Silva, C.R.d., Martins, E., Silveira, A.C.P., Simeão, M., Mendes, A.L.,
Musina, Olga, Putnik, Predrag, Koubaa, Mohamed, Barba, Perrone, Í.T., Schuck, P., de Carvalho, A.F., 2017.
Francisco J., Greiner, Ralf, Granato, Daniel, Roohinejad, Thermodynamic characterization of single-stage
Shahin, 2017. Application of modern computer algebra spray-dryers: mass and energy balances for milk drying. Dry
systems in food formulations and development: a case study. Technol. 35 (15), 1791–1798
Trends Food Sci. Technol. 64, 48–59, ISSN 0924-2244 http://dx.doi.org/10.1080/07373937.2016.1275675.
https://doi.org/10.1016/j.tifs.2017.03.011. Tonon, R.V., Brabet, C., Hubinger, M.D., 2008. Influence of process
Nath, S., Satpathy, G.R., 1998. A systematic approach for conditions on the physicochemical properties of açai (Euterpe
investigation of spray drying processes. Dry. Technol. 16 (6), oleraceae Mart.) powder produced by spray drying. J. Food Eng.
1173–1193. 88 (3), 411–418.
Normand, V., Subramaniam, A., Donnelly, J., Bouquerand, P.-E., Tontul, Ismail, Topuz, Ayhan, 2017. Spray-drying of fruit and
2013. Spray drying: thermodynamics and operating vegetable juices: effect of drying conditions on the product
conditions. Carbohydr. Polym. 97 (2), 489–495. yield and physical properties. Trends Food Sci. Technol. 63,
Oakley, D.E., 2004. Spray dryer modeling in theory and practice. 91–102, ISSN 0924-2244
Dry. Technol. 22 (6), 1371–1402. https://doi.org/10.1016/j.tifs.2017.03.009.
Park, C.W., Stout, M.A., Drake, M., 2016. The effect of spray-drying Vehring, R., 2008. Pharmaceutical particle engineering via spray
parameters on the flavor of nonfat dry milk and milk protein drying. Pharm. Res. 25 (5), 999–1022.
concentrate 70%. J. Dairy Sci. 99 (12), 9598–9610. Vicente, J., Pinto, J., Menezes, J., Gaspar, F., 2013. Fundamental
Prinn, K.B., Costantino, H.R., Tracy, M., 2002. Statistical modeling analysis of particle formation in spray drying. Powder
of protein spray drying at the lab scale. AAPS PharmSciTech 3 Technol. 247, 1–7.
(1), 32–39, http://dx.doi.org/10.1208/pt030104. Walzel, P., 2011. Influence of the spray method on product quality
Reddy, R.S., Ramachandra, C.T., Hiregoudar, S., Nidoni, U., Ram, J., and morphology in spray drying. Chem. Eng. Technol. 34 (7),
Kammar, M., 2014. Influence of processing conditions on 1039–1048, http://dx.doi.org/10.1002/ceat.201100051.
functional and reconstitution properties of milk powder made Woo, M., Bhandari, B., 2013. Spray Drying for Food Powder
from Osmanabadi goat milk by spray drying. Small Rumin. Production Handbook of Food Powders. Woodhead Publishing
Res. 119 (1), 130–137. Limited, Cambridge, UK, pp. 29–56.
Robinson, S., 2014. Simulation: The Practice of Model Woo, M.W., Daud, W.R.W., Mujumdar, A.S., Wu, Z., Meor Talib,
Development and Use. Palgrave Macmillan. M.Z., Tasirin, S.M., 2008. CFD evaluation of droplet drying
Saravacos, G.D., Maroulis, Z.B., 2001. Transport Properties of models in a spray dryer fitted with a rotary atomizer. Dry.
Foods. CRC Press. Technol. 26 (10), 1180–1198.
Schiffter, H., Lee, G., 2007. Single-droplet evaporation kinetics and Zbicinski, I., Strumillo, C., Delag, A., 2002. Drying kinetics and
particle formation in an acoustic levitator. Part 1: evaporation particle residence time in spray drying. Dry. Technol. 20 (9),
of water microdroplets assessed using boundary-layer and 1751–1768.
acoustic levitation theories. J. Pharm. Sci. 96 (9), 2274–2283.

You might also like