optimization of spray drying process in cheese powder production
TRANSCRIPT
food and bioproducts processing 9 3 ( 2 0 1 5 ) 156–165
Contents lists available at ScienceDirect
Food and Bioproducts Processing
j ourna l h omepage: www.elsev ier .com/ locate / fbp
Optimization of spray drying process in cheesepowder production
Zafer Erbaya, Nurcan Kocab,∗, Figen Kaymak-Ertekinb, Mustafa Ucuncub
a Department of Food Engineering, Faculty of Engineering and Natural Sciences, Adana Science andTechnology University, 01180 Adana, Turkeyb Department of Food Engineering, Faculty of Engineering, Ege University, 35100 Izmir, Turkey
a b s t r a c t
In this study, white cheese powder was produced using a pilot scale spray drier and response surface methodology was
used to optimize the operating conditions of spray drying. The independent variables were inlet drying temperature,
atomization pressure and outlet drying temperature, while drying experiments were carried out with an inlet drying
air temperature range of 160–230 ◦C, an outlet drying air temperature range of 60–100 ◦C and an atomization pressure
range of 294–588 kPa. The responses were nonenzymatic browning index, free fat content, solubility index, bulk
density of cheese powder and exergy efficiency of the spray drying process. Optimum operating conditions were
found to be an inlet drying temperature of 174 ◦C, atomization pressure of 354 kPa, and an outlet drying temperature◦
of 68 C. At this optimum condition, nonenzymatic browning index, free fat content, solubility index, bulk densityand exergy efficiency were found to be 0.123 OD/g dm, 40.7%, 82.7%, 252 kg/m3 and 4.81%, respectively.
© 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Keywords: Spray drying; Optimization; Cheese powder; Free fat; Exergy
feature for overall reconstitution quality and food powders
1. Introduction
Cheese is one of the most important dairy products with morethan 1000 different varieties (Fox, 2011). Apart from the directconsumption of cheese, it can be used as a food ingredient forits functional properties such as flavor delivery, mouthfeel,appearance and adhesive properties. To obtain these func-tional properties, cheese needs be processed and the mosteffective industrial process is drying (Fox et al., 2000; Guineeand Kilcawley, 2004; Guinee, 2011). One of the most importantdehydrated cheese products is cheese powder. Today, cheesepowder finds widespread use in industrial sectors primarily asa flavoring agent and/or nutritional supplement in a variety offoods, and recent market analyses indicate that the consump-tion of cheese as an ingredient is growing rapidly (Guinee,2011).
Drying of food materials is complicated because physical,
chemical and biochemical transformations may occur duringdrying, some of which may be desirable. So, in practice, a∗ Corresponding author at: Department of Food Engineering, Faculty
Tel.: +90 232 3113029; fax: +90 232 342 7592.E-mail addresses: [email protected] (Z. Erbay), nurcan.koca@e
(F. Kaymak-Ertekin), [email protected] (M. Ucuncu).Received 8 April 2013; Received in revised form 27 November 2013; AcAvailable online 2 January 2014
0960-3085/$ – see front matter © 2014 The Institution of Chemical Engihttp://dx.doi.org/10.1016/j.fbp.2013.12.008
dryer is considerably more complex than a device that merelyremoves moisture and for every dryer, the process conditionsmust be determined based on the feed, the product being pro-duced, the purpose of the drying and methods being employed(Mujumdar and Law, 2010).
Spray drying is a suspended particle processing techniquethat has become one of the most important methods for dryingfluid foods, especially in the dairy industry (Filkova et al., 2006).Dairy powders are generally characterized by their physicalproperties such as bulk density, reconstitution properties andfree fat content; and these properties are directly affected bythe spray drying process (Schuck, 2002).
Bulk density is one of the properties used as part of thespecifications for final product and it is used by the indus-try to arrange storage, processing, packaging and distributionconditions (Barbosa-Canovas et al., 2005). Solubility is a key
of Engineering, Ege University, 35100 Bornova, Izmir, Turkey.
ge.edu.tr, [email protected] (N. Koca), [email protected]
cepted 20 December 2013
should be able to provide good solubility to be useful andfunctional (Barbosa-Canovas et al., 2005; Baldwin and Truong,
neers. Published by Elsevier B.V. All rights reserved.
ssing
2watcdiMta1d
udidasaabiesE
opc
2
2
Wfa5bpc
wcpTLa8hd
2
TdsvpPadr
food and bioproducts proce
007). Additionally, free-fat that is defined as the fraction of fathich is not protected by protein film is very important during
nd after processing as it leads to off-flavors, poor rehydra-ion and flowing properties (Farkye, 2006). Apart from these,onsiderable quality concerns for powdered products are theegradation of flavor, color and texture of the product dur-
ng processing or storage. Nonenzymatic browning, known asaillard reaction, is one of the major detrimental reactions
hat causes the formation of chemically stable and nutrition-lly unavailable derivatives for dairy powders (Palombo et al.,984; Kilic et al., 1997), because of that, it should be minimizeduring spray drying.
Furthermore, drying is one of the most energy-intensivenit operations and the most energy-intensive method forrying is spray drying with the sole exception of freeze dry-
ng (Mujumdar and Huang, 2007). In the dairy industry, sprayrying operations are the primary drying operations and theyre the most important processes with regards to energy con-umption (IDF, 2005). An important tool to analyze, optimizend improve the energy efficiency of the spray drying processnd spray driers can be exergy analysis. Several studies haveeen undertaken on exergy analysis of different kinds of dry-
ng processes of food materials (Erbay and Icier, 2011; Gungort al., 2011), while studies focused on exergetic assessment ofpray drying process have recently begun (Jin and Chen, 2011;rbay and Koca, 2012; Aghbashlo et al., 2012).
The main objective of this study was to determine theptimum process conditions for spray drying in white cheeseowder production by investigating the effects of drying pro-ess variables on product quality.
. Materials and methods
.1. Materials
hite cheese that was ripened for 7 months was suppliedrom Sütas Dairy Company (Bursa, Turkey). Water, fat, protein,sh and salt content of white cheese used in this study were2.0%, 24.4%, 18.4%, 4.9% and 4.3%, respectively. White cheeselocks were ground into small pieces. The ground cheese wasut into air- and water-tight durable polypropylene plasticontainers, stored at 2 ◦C and processed within 48 h.
Before drying, white cheese slurry composed of groundhite cheese, water and Joha emulsifying salts (Kipa Chemi-
al Company, Istanbul, Turkey) was prepared. In cheese slurryreparation, 3% (based on cheese) emulsifying salts were used.he slurry was heated and sheared with a blender (modelB10S, Waring, Torrington, CT). Firstly, the slurry was shearedt 6000 rpm for 1 min. Then, slurry heated in a water bath to0 ◦C and sheared again at 6000 rpm for 10 min. The slurry,aving 25% dry matter based on cheese, was fed to the sprayrier at 45 ◦C.
.2. Drying procedure
he white cheese slurry was dried in a pilot scale sprayrier (Mobile Minor Niro-Atomizer, Soeborg, Denmark). Thechematic illustration of the spray drier was shown in a pre-ious study (Erbay and Koca, 2012). The cheese slurry wasumped with a peristaltic pump (model BT600-2J, Longerrecision Pump, Baoding, Hebei, China) to the atomizer and
tomized with a rotary atomizer into a drying cabinet withimensions of 1.2 m height and 0.87 m diameter. The feedate was adjusted due to the inlet and outlet drying air9 3 ( 2 0 1 5 ) 156–165 157
temperatures. The rotary atomizer was driven by an airturbine wheel and its speed was between 20,000 rpm and31,500 rpm. The atomizer wheel had a 50 mm diameter with24 vanes to create a perfect and uniform atomization. Atom-ization pressure affects the droplet dimensions sprayed intothe drying cabinet; therefore it is an important factor thatcauses effects on the drying rate and powder dimensions.Experiments were carried out at the inlet drying air tem-perature range of 160–230 ◦C, outlet drying air temperaturerange of 60–100 ◦C and with an atomization pressure range of294–588 kPa. The air flow was co-current and the mass flowrate of air was 0.08 kg/s. The cheese powder samples werepackaged in PET/Al/LDPE, than analyzed within 48 h. Beforestarting any experiment, the system was run for at least halfan hour to obtain steady-state conditions.
2.3. Moisture content
The moisture contents of white cheese powders were deter-mined by using gravimetric method (IDF, 1982). Two grams ofsamples were placed in an oven at 102 ◦C until constant weightwas obtained. Masses were measured using a digital balance(model UX4200H, Shimadzu, Kyoto, Japan).
2.4. Browning index
The browning index values of white cheese powder sampleswere measured by an enzymatic digestion method (Palomboet al., 1984; Kilic et al., 1997). The method was based onpronase proteolysis that releases the brown pigments. Onegram of white cheese powder was used during analysis andthe absorbance of the samples were read at 420 (A420) and 550(A550) nm by using a spectrophotometer (Varian, Cary 50 BioUV/Visible Spectrophotometer, Palo Alto, CA) and the opticaldensity (OD) of the samples were calculated as
OD = A420 − A550 (1)
The browning index (BI) values for each sample wereexpressed as OD/g dry matter.
2.5. Free fat content
The free fat content of dairy powders is commonly quantifiedby the solvent extraction method (Vignolles et al., 2007). Inthis study, the modified solvent extraction method describedin A/S Niro Atomizer (2005) was used and the results wereexpressed as the percentage of free-fat content.
2.6. Solubility index
The solubility index of white cheese powder samples weremeasured by the Haenni method (Hawthorne, 1944). 1 g ofpowder was placed into a stoppered test tube with 5 mL of5% sodium chloride solution (m/v). The solution was shakenand mixed with a vortex mixer. The refractive index valuesof the dispersed sample and sodium chloride solution weremeasured by using a refractometer (model RFM 330, Belling-ham Stanley Ltd., Kent, UK). The solubility index (SI) valueswere calculated as:
Haenni value = y = (RIsample − RINaCl) × 1000 (2)
cessi
158 food and bioproducts proSI = log(y) − 0.4450.01
(3)
2.7. Bulk density
The bulk density of powders is commonly measured by weigh-ing the mass of the powder that was poured into a vessel ofknown volume (Barbosa-Canovas et al., 2005). In this study,white cheese powder was gently poured from a specifiedheight into a 100 mL tared graduated measuring cylinder tothe 100 mL mark and the weight of the powder was recorded.The bulk density of the white cheese powder was calculatedfrom the recorded mass/marked volume ratio (Jinapong et al.,2008).
2.8. Exergy efficiency
Exergy is defined as the maximum amount of work and is ameasure of the potential of a stream to cause change (Dincerand Sahin, 2004). Exergy analysis of the spray drying processwas performed according to the method described by Erbayand Koca (2012) and the exergy efficiency of the spray dryingprocess was used as an optimization response. Exergy effi-ciency is defined as the ratio of total exergy out to total exergyin where “out” refers to “net output” or “product” or “desiredvalue”, and “in” refers to “given” or “used” or “fuel” (Dincer andSahin, 2004).
The measurements were taken to perform the exergyanalysis of the system during the drying process. Air humidi-ties, temperatures and velocities at the inlet and outletof the spray drier were measured with robust humid-ity probes (model 0636.2140, Testo, Freiburg, Germany),vane/temperature probes (model 0635.9540, Testo) and pro-fessional telescopic handle for plug-in vane probes (model0430.0941, Testo), respectively. Measurements of the dryingair temperature, velocity and relative humidity were recordedevery 2 min. The ambient temperature and relative humiditywere also measured and recorded. The surface temperatureof the drying ducts was measured with surface temperatureprobes (model 0628.0019, Testo). Power consumption of thesystem was measured with an electricity meter and electriccurrent–voltage values of system components were measuredwith a clamp meter (model EX730, Extech, Nashua, NH). All themeasurements were simultaneously observed and recordedwith a multi-function instrument (350-XL/454, Testo) and log-gers.
2.9. Experimental uncertainty
Uncertainty analysis is needed to prove the accuracy of theexperiments especially experimental results that are mainlybased on instrumental measurements as done for exergyanalysis in this study. In the present study, temperatures,relative humidities, air velocities, mass losses, electrical con-sumption parameters and drying times were measured withappropriate instruments explained mentioned above, andtotal uncertainties for all these measurements were calcu-lated individually. The accuracy of temperature measuringequipment was ±0.2 ◦C and reading error for temperaturemeasurements was assumed as ±0.1 ◦C. The accuracy of dig-ital balance was ±0.01 g and reading error was assumed as
±0.01 g. The accuracy of the velocity probes used in air veloc-ity measurements was ±0.2 m/s and the error coming fromthe flow disorder was assumed as ±0.05 m/s. The accuracyng 9 3 ( 2 0 1 5 ) 156–165
of relative humidity probes was ±2% RH and reading errorswere assumed as ±0.1 RH. Furthermore, the accuracy ofthe surface temperature measurement probes accuracies was±0.1 ◦C. While the accuracy of the clamp meter was ±0.2 Aand reading error was assumed as ±0.1 A for electric currentmeasurements, they were ±10 and ±10 V for electric poten-tial difference measurements, respectively. Additionally, therewere reading and timing errors during time productions thatwere assumed as 2 and 30 s, respectively. According to all theseuncertainties and errors, a detailed uncertainty analysis wascarried out using the method described by Holman (2001) forthe experimental measurements of the thermal propertiesand the total uncertainties of the predicted values for exergyefficiency values:
UF =[(
∂F
∂z1u1
)2+(
∂F
∂z2u2
)2+ . . . +
(∂F
∂znun
)2]1/2
(4)
2.10. Experimental design, statistical analysis andoptimization
Response surface methodology (RSM) was used to investi-gate the main effects of spray drying process variables onthe browning index (BI), free fat content (FFC), solubility index(SI), bulk density (BD) and exergy efficiency (ε) during produc-tion of white cheese powder. Inlet drying temperature (x1,Tin), atomization pressure (x2, P) and outlet drying tempera-ture (x3, Tout) were selected as independent variables. Processvariable ranges were determined by the help of the literaturesurvey (Guinee and Kilcawley, 2004; Kumar and Mishra, 2004;Koc et al., 2010; Koc et al., 2011) and by means of preliminaryexperiments. During the preliminary experiments, it is aimedto produce cheese powder with moisture content of 5% orlower. A central composite rotatable design (CCRD) including20 experiments formed by 6 central points and 6 (� = 1.68179)axial points to 23 full factorial design was used (Table 1).
Experimental data were fitted to a second-order polyno-mial model and regression coefficients were obtained foreach response. Significant terms in the models were foundby analysis of variance (ANOVA) and significance was judgedby the F-statistic calculated from the data. Model adequa-cies were checked by R2, adj-R2, pre-R2, Adeq. Precision,PRESS and C.V. (lack of fit > 0.1; R2 > 0.94; (Adj-R2 − Pre-R2) < 0.2;max.PRESS; C.V. < 10; Pre-R2 > 0.7; Adeq. Precision > 4) (Myersand Montgomery, 2002). After model fitting, residual analysesincluding the examination of diagnostic plots and calculationof case statistics were conducted to validate assumptions usedin ANOVA. Design Expert Ver. 7.0.0 (Stat-Ease, 2005) was usedto fit response surfaces and optimize the drying process.
The desirability function method was used to optimizemultiple responses, simultaneously. In this method, a desir-ability function which reflects the desirable ranges (0–1) foreach response (di) should be generated. Furthermore, animportance term (ri) (from 1 to 5) can be assigned for eachresponse due to the importance for optimization (Eren andKaymak-Ertekin, 2007). The final desirability function equa-tion is shown below:
D(x) = (dri1 × dr2
2 × dr33 × dr4
4 × dr55 )1/
∑ri =(
5∏i=1
drii
)1/∑
ri
(5)
In the present study, the cheese powder which had theminimum color variation due to the heat treatment, the least
food and bioproducts processing 9 3 ( 2 0 1 5 ) 156–165 159
Table 1 – Experimental design and data of response variables for central composite rotatable design.
Run # Tin (◦C) P (kPa) Tout (◦C) BI (OD/g dm) FFC (%) SI (%) BD (kg/m3) ε (%)
1 195 (0) 588(+1.68) 80 (0) 0.142 ± 0.005 38.7 ± 0.2 74.8 ± 1.9 258 ± 4 3.61 ± 0.142 216 (+1) 528 (+1) 92 (+1) 0.227 ± 0.005 37.1 ± 0.2 70.6 ± 1.2 227 ± 6 3.61 ± 0.143 216 (+1) 354 (−1) 68 (−1) 0.140 ± 0.001 41.4 ± 0.4 79.2 ± 0.6 246 ± 4 6.71 ± 0.284 174 (−1) 529 (+1) 68 (−1) 0.173 ± 0.003 40.8 ± 0.4 73.3 ± 0.7 261 ± 8 3.62 ± 0.145 195 (0) 441 (0) 80 (0) 0.117 ± 0.007 41.2 ± 0.3 77.1 ± 0.7 232 ± 6 4.45 ± 0.186 174 (−1) 529 (+1) 92 (+1) 0.213 ± 0.005 37.6 ± 0.4 62.0 ± 2.7 234 ± 7 1.84 ± 0.077 195 (0) 441 (0) 80 (0) 0.130 ± 0.004 41.6 ± 0.1 77.0 ± 1.5 233 ± 7 4.28 ± 0.178 195 (0) 294 (−1.68) 80 (0) 0.188 ± 0.009 40.0 ± 0.7 78.9 ± 1.4 247 ± 6 5.69 ± 0.239 216 (+1) 529 (+1) 68 (−1) 0.128 ± 0.003 42.6 ± 1.0 77.6 ± 1.3 251 ± 8 5.69 ± 0.24
10 216 (+1) 354 (−1) 92 (+1) 0.294 ± 0.007 39.9 ± 1.3 63.8 ± 2.0 229 ± 2 4.92 ± 0.2011 195 (0) 441 (0) 100 (+1.68) 0.337 ± 0.002 35.9 ± 0.2 58.4 ± 2.2 215 ± 2 2.84 ± 0.1112 195 (0) 441 (0) 80 (0) 0.115 ± 0.003 40.9 ± 0.2 76.6 ± 0.5 237 ± 9 4.47 ± 0.1813 195 (0) 441 (0) 80 (0) 0.115 ± 0.005 40.5 ± 0.3 74.8 ± 1.9 232 ± 8 4.30 ± 0.1714 174 (−1) 354 (−1) 92 (+1) 0.223 ± 0.012 38.7 ± 0.2 68.1 ± 2.3 238 ± 4 2.68 ± 0.1115 195 (0) 441 (0) 60 (−1.68) 0.156 ± 0.004 42.0 ± 0.3 78.8 ± 0.9 244 ± 6 5.76 ± 0.2416 195 (0) 441 (0) 80 (0) 0.120 ± 0.003 40.4 ± 0.1 75.8 ± 3.0 241 ± 7 4.41 ± 0.1817 160 (−1.68) 441 (0) 80 (0) 0.129 ± 0.002 39.7 ± 0.3 72.8 ± 1.0 255 ± 3 2.62 ± 0.1018 195 (0) 441 (0) 80 (0) 0.125 ± 0.008 40.8 ± 0.3 74.4 ± 2.1 235 ± 3 4.67 ± 0.1819 174 (−1) 354 (−1) 68 (−1) 0.120 ± 0.002 40.7 ± 0.2 82.1 ± 1.0 251 ± 2 4.70 ± 0.19
8 ± 0
sbtddiaMo
3
Rsfa0rdBttafswrsaF
ySebwcoTnl
20 230(+1.68) 441 (0) 80 (0) 0.13
ensitivity to oxidation, the highest ability to solubilize, theest quality for packaging and distribution and the produc-ion which had the lowest energy inefficiencies during sprayrying were aimed. Therefore, the desirability functions wereeveloped for the criteria: minimum BI and FFC, and max-
mum SI, BD and ε. Additionally, the same importance forll responses (r = 3) was appointed (Erbay and Icier, 2009a,b).oreover, one sample T-test was performed for verification of
ptimization.
. Results and discussion
esults of different runs of spray drying experiments arehown in Table 1. The moisture contents of powders wereound between 1.35 and 2.97%. The ranges of standard devi-tions for the results of BI, FFC, SI and BD were stated as.001–0.012 OD/g dry matter, 0.1–1.3%, 0.5–3.0% and 2–9 kg/m3,espectively. Furthermore, the highest proportions of standardeviations to mean values were 6.5%, 3.2%, 4.4% and 3.7% forI, FFC, SI and BD, respectively. Uncertainties of the experimen-al measurements were shown in Table 2 and the ranges ofotal uncertainties for exergy efficiency values were obtaineds 3.82–4.19%. The maximum uncertainty values obtainedrom exergetic efficiencies were under 5% which is rea-onable for an experimental study (Granström, 2005). CCRDhich includes 6 central points that are utilized to show the
eproducibility of productions was used in this study. Thetandard deviations at these central points were calculateds 0.006 OD/g dry matter, 0.4%, 1.5%, 1 kg/m3 and 0.14% for BI,FC, SI, BD and ε, respectively.
Multiple linear regression analysis of the experimental dataielded second order polynomial models for predicting BI, FFC,I, BD and ε. ANOVA was conducted to determine significantffects of process variables on each response (Table 3). Effectseing not significant (p > 0.05) were stepped down from modelsithout damaging the model hierarchy. Regression equation
oefficients of the proposed models and statistical significancef all main effects calculated for each response are shown in
able 4. ANOVA also showed that the lack of fit was not sig-ificant for any response surface models at a 95% confidenceevel and that model adequacies were appropriate (Table 4).
.006 41.9 ± 0.3 75.7 ± 1.1 238 ± 5 5.55 ± 0.22
To visualize the general trends of the variations of responsesvia process parameters a response optimizer plot was drawn(Fig. 1) and to visualize the interaction effects of two factorson any response, the response surface and contour plots weredrawn for each of the fitted models as the function of twoindependent variables, while keeping the other variable at thecentral value (Figs. 2–8). Effects of variables on responses weredetermined by evaluation of these plots. Although the mostimportant process variable was seemed to be the outlet dry-ing air temperature (Fig. 1), the discussion about the effectsof variables on responses was deepened in the followingsections.
3.1. Browning index (BI)
All process factors discussed in this study and their linear,squared and interaction terms were determined to be signifi-cant for BI values (Table 4). However, the importance of processfactors for BI can be prioritized in descending order as: out-let drying temperature, atomization pressure and inlet dryingtemperature. The surface plot of BI vs inlet air drying temper-ature and atomization pressure is shown in Fig. 2. To obtainlow BI values, higher atomization pressures should be usedfor high inlet drying temperatures and moderate atomizationpressures should be chosen for low inlet drying temperatures.In other words, the atomization pressure affected BI val-ues more than inlet drying temperature and low atomizationpressures should be avoided. The response surface plot for thevariation of BI values with inlet–outlet drying temperaturesshowed that outlet drying temperature had a significant effecton BI values (Fig. 3). BI values increased sharply, especiallywith higher than 85 ◦C outlet drying temperatures. Fig. 4 showsthe interaction effects of outlet drying temperature and atom-ization pressure on BI. There were clear curvatures and anoptimum field revealed for minimum BI. Briefly, the responsesurface plots showed that outlet drying temperatures less than80 ◦C with medium atomization pressures should be used toobtain low BI values. It’s well known that the Maillard reaction
is induced by thermal effect (Hardy et al., 1999). Experimentaldata showed this thermal effect could be clearly seen for highoutlet drying temperatures.
160 food and bioproducts processing 9 3 ( 2 0 1 5 ) 156–165
Table 2 – Uncertainties of the experimental measurements.
Experimental measurements Unit Comment
Uncertainty in the temperature measurement ◦C ± 0.224Uncertainty in the weight measurement g ± 0.014Uncertainty in the air velocity measurement m/s ± 0.21Uncertainty in the measurement of relative humidity of air % ± 0.41Uncertainty in the surface temperature measurement ◦C ± 0.1Uncertainty in the electric current measurement A ± 0.224Uncertainty in the electric potential difference measurement V ± 14.14Uncertainty in the time measurement s ± 30.1
Table 3 – ANOVA evaluation of linear, quadratic and interaction terms for each response variables (x1: Tin; x2: P; x3: Tout).
Source DF BI FFC SI BD ε
Sum ofsquares
p-Value Sum ofsquares
p-Value Sum ofsquares
p-Value Sum ofsquares
p-Value Sum ofsquares
p-Value
Model 9 0.075 <0.0001 54.70 <0.0001 707.14 <0.0001 2358.90 <0.0001 36.60 <0.0001x1 1 0.00042 0.0326 3.47 0.0055 8.17 0.0341 257.11 0.0006 15.82 <0.0001x2 1 0.00094 0.0041 1.77 0.0304 19.64 0.0035 58.15 0.0399 4.58 <0.0001x3 1 0.036 <0.0001 36.70 <0.0001 491.33 <0.0001 1224.44 <0.0001 15.60 <0.0001x1x2 1 0.00186 0.0004 0.048 0.6881 50.34 0.0001 2.74 0.6195 0.0072 0.6719x1x3 1 0.00148 0.0009 0.41 0.2522 1.02 0.4072 0.020 0.9659 0.0001 0.9609x2x3 1 0.00176 0.0005 3.40 0.0058 15.30 0.0073 57.46 0.0408 0.00679 0.6808x2
1 1 0.00043 0.0319 0.00552 0.8910 9.01 0.0276 203.42 0.0013 0.45 0.0062x2
2 1 0.00402 <0.0001 3.57 0.0050 0.21 0.6999 477.83 <0.0001 0.032 0.3811x2
3 1 0.03 <0.0001 5.93 0.0010 114.44 <0.0001 68.46 0.0283 0.11 0.1202Residual 10 0.00069 2.79 13.58 104.27 0.38
Lack of fit 5 0.00051 0.1414 1.77 0.2776 6.87 0.4906 39.25 0.7035 0.24 0.2710Pure error 5 0.00018 1.02 6.72 65.03 0.14
Total 19 0.075 57.49 720.72 2463.18 36.98
R2 0.9909 0.9515 0.9812 0.9577 0.9905Adj-R2 0.9826 0.9078 0.9642 0.9196 0.9819Pre-R2 0.9450 0.7097 0.9053 0.8365 0.9469Adeq.
precision36.537 16.096 30.487 21.229 41.550
PRESS 0.00414 16.69 68.24 402.82 1.55
C.V. 4.98 1.323.2. Free fat content (FFC)
Few recent studies have been published about high-fat dairypowders and very few of them focused on the effects of dryingon free fat contents (Vignolles et al., 2007). Kelly et al. (2002)
Table 4 – ANOVA evaluation for each response variable and coeinsignificant factors from the models (x1: Tin; x2: P; x3: Tout).
Source BI FFC SI
Coefficient p-Value Coefficient p-Value Coeffi
Model 0.12 <0.0001 40.90 <0.0001 76.0x1 0.00554 0.0326 0.5 0.0026 0.7x2 −0.00829 0.0041 −0.36 0.0197 −1.2x3 0.051 <0.0001 −1.63 <0.0001 −5.9x1x2 −0.015 0.0004 2.4x1x3 0.013 0.0009x2x3 −0.015 0.0005 −0.65 0.0028 1.3x2
1 0.00542 0.0319 −0.8x2
2 0.017 <0.0001 −0.5 0.0022
x23 0.045 <0.0001 −0.64 0.0003 −2.8Lack of fit 0.1414 0.3768
R2 0.9909 0.9434 0.9Adj-R2 0.9826 0.9172 0.9Pre-R2 0.9450 0.8526 0.9Adeq. precision 36.537 19.739 35.0PRESS 0.00414 8.48 44.1C.V. 4.98 1.25 1.5
1.58 1.34 3.86
reported that the rise in the fat content of dairy powders inten-sively increased FFC values. Increasing the fat content from30% to 70% caused an increase in FFC values from 5% to 60%
(Kelly et al., 2002). In this study, the average fat content ofcheese powders was approximately 46.7% and FFC values werefficient of prediction models after removing the
BD ε
cient p-Value Coefficient p-Value Coefficient p-Value
8 <0.0001 235.24 <0.0001 4.42 <0.00017 0.0244 −4.32 0.0002 0.95 <0.0001
0.0018 2.06 0.0253 −0.57 <0.00017 <0.0001 −9.42 <0.0001 −0.92 <0.00019 <0.0001
7 0.0042 −2.65 0.0260 0.0175 3.74 0.0005 −0.15 0.0048
5.75 <0.00012 <0.0001 −2.17 0.0169
0.5873 0.8292 0.3069
794 0.9565 0.9854675 0.9312 0.9815387 0.8842 0.969186 25.380 58.1188 285.25 0.901 1.24 3.90
food and bioproducts processing 9 3 ( 2 0 1 5 ) 156–165 161
Fig. 1 – General trends of the variations of responses via process parameters. Central values used (Ti = 195 ◦C, P = 441 kPa,T
otK
t(lHc2d
out = 80 ◦C) while calculating the process parameters.
btained in the range of 35.9–42.6%, which was consistent withhe literature (Table 2) (Kelly et al., 2002; Vignolles et al., 2007;im et al., 2009).
It has been reported that there are conflicting experimen-al results about the effect of inlet drying temperature on FFCVignolles et al., 2007). In this study, there was a linear corre-ation with inlet drying temperature and FFC values (Table 4).igher inlet drying temperature increased the formation of
apillaries and vacuoles that make fat unprotected (Farkye,006) and this resulted in higher FFC values at higher inletrying temperatures.Linear, squared and interaction terms of atomization pres-sure and outlet drying temperature for FFC values wereobtained as significant and inversely affected FFC (Table 4).Table 4 shows that linear and squared term for atomizationpressure showed that an increase in atomization pressurecaused a decrease in FFC. It is likely that atomization enhancedthe homogenization effect of spraying which promoted theemulsifying of fat in the cheese slurry and decreased FFC (Kelly
et al., 2002; Vignolles et al., 2007). Furthermore, it is reportedthat FFC is inversely related to particle size and rising atomiza-tion pressure causes the decrease of the particle size (Farkye,
162 food and bioproducts processing 9 3 ( 2 0 1 5 ) 156–165
Fig. 2 – Response surface and contour plot for BI at constantoutlet temperature (80 ◦C).
Fig. 5 – Response surface and contour plot for FFC atconstant inlet temperature (195 ◦C).
Fig. 6 – Response surface and contour plot for SI at constant
Fig. 3 – Response surface and contour plot for BI at constantatomization pressure (441 kPa).2006). Fig. 5 shows the interaction effect of atomization pres-sure and outlet drying temperature on the variation of FFC. Itis clear that to obtain the lowest FFC values, atomization pres-sure should be increased, especially at higher outlet dryingtemperatures (Fig. 5).
3.3. Solubility index (SI)
All process factors and their linear, squared and interactionterms, except from the squared term of atomization pressure
and interaction effect of inlet–outlet drying temperature, wereobtained as significant for SI values (Table 4). However, theFig. 4 – Response surface and contour plot for BI at constantinlet temperature (195 ◦C).
inlet temperature (195 ◦C).
most important process factor was outlet drying temperature.Increasing the outlet drying temperature caused a decreasein SI and it can be clearly seen from Fig. 6. These resultsare in harmony with the results reported in the literature(Anandharamakrishnan et al., 2007, 2008; Bakar et al., 2013).On the other hand, atomization pressure affected less effec-tively (Fig. 6). The powders obtained from productions madewith lower atomization pressures were more soluble (Table 4).Fig. 7 shows the interaction effect of inlet drying temperatureand atomization pressure on SI. Both higher inlet drying tem-
peratures with higher atomization pressures and lower inletdrying temperatures with lower atomization pressures shouldFig. 7 – Response surface and contour plot for SI at constantoutlet temperature (80 ◦C).
food and bioproducts processing
Fc
bu
3
GaaisiBtBlmptgfl
adettt
ig. 8 – Response surface and contour plot for BD atonstant inlet temperature (195 ◦C).
e chosen as operation parameters for obtaining higher SI val-es (Fig. 7).
.4. Bulk density (BD)
enerally, average BD values of food powders vary between 300nd 800 kg/m3and BD of skim milk powders vary between 400nd 450 kg/m3 (Kelly et al., 2002). The main properties affect-ng BD values are the powders’ structure and fat content. Amooth and uniform structure increases BD, whereas poros-ty that increases the air in the powder structure decreasesD (Kelly et al., 2002). Furthermore, the rise in the fat con-ent causes a decrease in BD (Kelly et al., 2002). In this study,D values were in the range of 215–261 kg/m3 (Table 1) were
ower than values from compared literature and one of theain reasons for this was the high fat contents of cheese
owders. Moreover, single stage spray drying was used whichends to give smaller particle sizes than when using an inte-rated fluidized bed, leading to increased cohesion and lowerowability.
Linear and squared effects of all process variables anddditionally, interaction effect of atomization pressure–outletrying temperature were important for the BD (Table 4). Lin-ar effects showed that the decrease in inlet and outlet drying
emperatures caused an increase in BD. On the other hand,he linear effect of atomization pressure was less impor-ant (Table 4). Furthermore, experimental results proved thatTable 5 – The predicted and experimental values for responses
Run # BI (OD/g dm) FFC (%)
1 0.127 ± 0.001 41.0 ± 0.2
2 0.129 ± 0.002 41.2 ± 0.1
3 0.120 ± 0.002 40.7 ± 0.3
Predicted values 0.123 40.7
Table 6 – Results of statistical analysis for verification of optimi
Response Predicted values Experimental valuesa
BI (OD/g KM) 0.123 0.125 (±0.005)
FFC (%) 40.7 41.0 (±0.3)
SI (%) 82.7 80.9 (±1.1)
BD (kg/m3) 251.6 249.6 (±0.9)
ε (%) 4.81 4.67 (±0.03)
a Experimental values were expressed as mean ± standard deviation.b Mean standard error.c The % error = (|yexp − ypre|/yexp) × 100.
9 3 ( 2 0 1 5 ) 156–165 163
low inlet–outlet drying temperatures with higher atomizationpressures should be used to obtain higher BD values (Fig. 8).Generally, low outlet drying temperatures promote uniformdrying of droplets, controlled particle shrinkage and resultedin higher BD (Kelly et al., 2002). However, higher inlet dryingtemperatures cause the formation of vacuoles and air bubbleswhich could decrease BD (Kelly et al., 2002; Farkye, 2006).
3.5. Exergy efficiency (ε)
The exergy efficiency was affected by linear effects of allprocess factors and squared effect of inlet drying temper-ature (Table 4). The exergy efficiency of the drying processwas improved by decreasing outlet drying temperature andatomization pressure and by increasing inlet drying temper-ature. These tendencies were similar to the results reportedin the studies focused on the energy efficiency of spray dryingprocess. The rise in the difference of inlet–outlet drying tem-perature (Al-Mansour et al., 2011) and/or the increase in thefeed flow rate (Kurozawa et al., 2011) caused an increase in thespray drying process performance.
3.6. Optimization
The optimization of the spray drying process in white cheesepowder production was conducted for selected ranges of inletdrying temperature, atomization pressure and outlet dryingtemperature as 160–230 ◦C, 294–588 kPa, and 60–100 ◦C, respec-tively. Desirability functions were developed for the criteria:minimum BI and FFC, and maximum SI, BD and ε. By applyingthe desirability function method, the results obtained for theoptimum spray drying conditions are: 174 ◦C for inlet dryingtemperature, 354 kPa for atomization pressure, and 68 ◦C foroutlet drying temperature. At this optimized process condi-tion, BI, FFC, SI, BD, and ε were calculated as 0.123 OD/g dm,40.7%, 82.7%, 252 kg/m3 and 4.81%, respectively.
Three productions were made for the verification of thepredicted models at the optimum process conditions. Theresults (BI, FFC, SI, BD, and ε) obtained from these verifica-tion productions were 0.125 ± 0.005 OD/g dm, 41.0 ± 0.3%,
80.9 ± 1.1%, 250 ± 1 kg/m3 and 4.67 ± 0.03%, respectively(Table 5). The results of statistical analysis for verifica-tion are shown in Table 6. No significant differences wereat optimum process condition.
SI (%) BD (kg/m3) ε (%)
80.0 ± 1.7 249 ± 1 4.6880.6 ± 2.4 249 ± 1 4.6482.1 ± 1.0 251 ± 2 4.70
82.7 252 4.81
zation.
SEb Difference % errorc p-Value
0.003 −0.002 1.86 0.4830.145 −0.267 0.65 0.2081.625 1.843 2.28 0.0980.517 2.033 0.81 0.0590.018 0.137 2.92 0.016
cessi
164 food and bioproducts prodetermined between experimental values and predictedvalues from optimization (p > 0.05), except for ε (Table 6).The difference and percentage error values for ε were 0.137and 2.92%, respectively. These values are small enoughand would be acceptable for practical approaches in ε.Briefly, the results of the verification productions were closeto the predicted values obtained from the optimizationmodels.
4. Conclusions
The process factors which were inlet and outlet drying tem-peratures with atomization pressure and the responses whichwere browning index, free fat content, solubility index, bulkdensity of product and exergy efficiency of the process wereused in order to optimize the spray drying process of whitecheese powder.
The main conclusions drawn from the results of thepresent study are as follows:
1. Outlet drying temperature was the most significant processparameter that affected BI and outlet drying temperaturesless than 80 ◦C with 350–520 kPa of atomization pressuresshould be used to obtain low BI values.
2. For low FFC values, low inlet and high outlet drying temper-atures with high atomization pressure should be chosen asoperation parameters.
3. The most significant process parameter for SI was outletdrying temperature, followed by atomization pressure andthe powders produced at lower outlet drying temperaturesand lower atomization pressures were more soluble.
4. Experimental results proved that the optimum processconditions to produce cheese powder with higher BD wereat the lowest inlet–outlet drying temperatures with thehighest atomization pressure.
5. To improve the exergy efficiency of a spray drying process,inlet drying temperature should be increased and outletdrying temperature and atomization pressure should bedecreased.
6. With an inlet drying temperature of 174 ◦C, an outlet dry-ing temperature of 68 ◦C, and an atomization pressure of354 kPa, minimum BI and FFC, and maximum SI, BD and ε
values were calculated. At this optimized point, BI, FFC, SI,BD, and ε were calculated as 0.123 OD/g dm, 40.7%, 82.7%,252 kg/m3 and 4.81%, respectively.
Acknowledgements
This study is a part of PhD thesis named as “Optimizationof Spray Drying and Determination of Effects of Using Wheyand Maltodextrin during White Cheese Production on Prod-uct Quality and Storage Stability”. The authors are grateful forthe financial support provided for the project no: 109O093 byThe Scientific and Technological Research Council of Turkey(TUBITAK) and for the project no: 2010/Bil/012 by Ege Univer-sity Science and Technology Center (EBILTEM), Sütas DairyCompany for providing cheese, Yildiz Group for providingatomizer and Kipa Chemical Company for providing emul-sifying salt. The authors would like to thank the reviewers
of the journal due to their valuable and constructive com-ments, which have been utilized to improve the quality of thepaper.ng 9 3 ( 2 0 1 5 ) 156–165
References
A/S Niro Atomizer, 2005. Determination of surface free fat ofpowder. In: GEA Niro Method No. A10a. A/S Niro Atomizer,Copenhagen http://www.niro.com/NIRO/CMSDoc.nsf/webdoc/ndkw6u9c45 (accessed on 28.01.13).
Aghbashlo, M., Mobli, H., Madadlou, A., Rafiee, S., 2012. Influenceof spray dryer parameters on exergetic performance ofmicroencapsulation process. Int. J. Exergy 10 (3), 267–289.
Al-Mansour, H.E., Al-Busairi, B.H., Baker, C.G.J., 2011. Energyconsumption of a pilot-scale spray dryer. Dry Technol. 29 (16),1901–1910.
Anandharamakrishnan, C., Rielly, C.D., Stapley, A.G.F., 2007.Effects of process variables on the denaturation of wheyproteins during spray drying. Dry Technol. 25 (5), 799–807.
Anandharamakrishnan, C., Rielly, C.D., Stapley, A.G.F., 2008. Lossof solubility of �-lactalbumin and �-lactoglobulin during thespray drying of whey proteins. LWT 41, 270–277.
Baldwin, A.J., Truong, G.N.T., 2007. Development of insolubility indehydration of dairy milk powders. Food Bioprod. Process. 85(C3), 202–208.
Bakar, J., Ee, S.C., Muhammad, K., Hashim, D.M., Adzahan, N.,2013. Spray-drying optimization for red pitaya peel (Hylocereuspolyrhizus). Food Bioprocess Technol. 6, 1332–1342.
Barbosa-Canovas, G.V., Ortega-Rivas, E., Juliano, P., Yan, H., 2005.Food Powders: Physical Properties, Processing, andFunctionality. Kluwer Academic/Plenum Publishers, New York(Chapter 3).
Dincer, I., Sahin, A.Z., 2004. A new model for thermodynamicanalysis of a drying process. Int. J. Heat Mass Transfer 47,645–652.
Erbay, Z., Icier, F., 2009a. Optimization of hot air drying of oliveleaves using response surface methodology. J. Food Eng. 91 (4),533–541.
Erbay, Z., Icier, F., 2009b. Optimization of drying of olive leaves ina pilot-scale heat pump dryer. Dry Technol. 27 (3), 416–427.
Erbay, Z., Icier, F., 2011. Energy and exergy analyses on drying ofolive leaves (Olea europaea L.) in tray drier. J. Food Process Eng.34 (6), 2105–2123.
Erbay, Z., Koca, N., 2012. Energetic, exergetic and exergoeconomicanalyses of spray drying process during white cheese powderproduction. Dry Technol. 30 (4), 435–444.
Eren, I., Kaymak-Ertekin, F., 2007. Optimization of osmoticdehydration of potato using response surface methodology. J.Food Eng. 79, 344–352.
Farkye, N., 2006. Significance of milk fat in milk powder. In: Fox,P.F., McSweeney, P.L.H. (Eds.), Advanced Dairy Chemistry;Volume-2 Lipids. , third ed. Springer, New York, pp.451–462.
Filkova, I., Huang, L.X., Mujumdar, A.S., 2006. Industrial spraydrying systems. In: Mujumdar, A.S. (Ed.), Handbook ofIndustrial Drying. , third ed. CRC Press, Boca Raton, FL, pp.215–254.
Fox, P.F., 2011. Cheese: overview. In: Fuquay, J.W., Fox, P.F.,McSweeney, P.L.H. (Eds.), Encyclopedia of Dairy Sciences:Volume 1. , second ed. Elsevier/Academic Press, London, pp.534–543.
Fox, P.F., Cogan, T.M., Guinee, T.P., McSweeney, P.L.H., 2000.Fundamentals of Cheese Science. Aspen Publication,Gaithersburg, MD (Chapter 19).
Granström, K., 2005. A method to measure emissions from dryerswith diffuse leakages 2 – Sensitivity studies. Dry Technol. 23(5), 1127–1140.
Guinee, T.P., 2011. Cheese as a food ingredient. In: Fuquay, J.W.,Fox, P.F., McSweeney, P.L.H. (Eds.), Encyclopedia of DairySciences: Volume 1. , second ed. Elsevier/Academic Press,London, pp. 822–833.
Guinee, T.P., Kilcawley, K.N., 2004. Cheese: cheese as aningredient. In: Fox, P.F., McSweeney, P.L.H., Cogan, T.M.,Guinee, T.P. (Eds.), Cheese Chemistry, Physics and
Microbiology. , third ed. Elsevier Academic Press, London, pp.395–428.
ssing
G
H
H
H
I
I
J
J
K
K
K
food and bioproducts proce
ungor, A., Erbay, Z., Hepbasli, A., 2011. Exergetic analysis andevaluation of a new application gas engine heat pumps(GEHPs) for food drying processes. Appl. Energy 88 (3),882–891.
ardy, J., Parmenties, M., Fanni, J., 1999. Functionality of nutrientsand thermal treatments of food. Proc. Nutr. Soc. 58 (3),579–585.
awthorne, J.R., 1944. Methods for determination of solubility ofdried whole egg. J. Soc. Chem. Ind. 63, 6–8.
olman, J.P., 2001. Analysis of experimental data. In: Holman, J.P.(Ed.), Experimental Methods for Engineers. , seventh ed.McGraw Hill, Singapore, pp. 48–143.
DF, 1982. Determination of the total solid content (cheese andprocessed cheese). In: IDF Standard 4A. International DairyFederation, Brussels, Belgium.
DF, 2005. Energy use in dairy processing. In: Bulletin of theInternational Dairy Federation. International DairyFederation, 401/2005.
in, Y., Chen, X.D., 2011. Entropy production during the dryingprocess of milk droplets in an industrial spray dryer. Int. J.Therm. Sci. 50 (4), 615–625.
inapong, N., Suphantharika, M., Jamnong, P., 2008. Production ofinstant soymilk powders by ultrafiltration, spray drying andfluidized bed agglomeration. J. Food Eng. 84, 194–205.
elly, J., Kelly, P.M., Harrington, D., 2002. Influence of processingvariables on the physicochemical properties of spray driedfat-based milk powders. Dairy Sci. Technol. 82, 401–412.
ilic, M., Muthukumarappan, K., Gunasekaran, S., 1997. Kineticsof nonenzymatic browning in cheddar cheese powder duringstorage. J. Food Process. Pres. 21 (5), 379–393.
im, E.H.J., Chen, X.D., Pearce, D., 2009. Surface composition of
industrial spray-dried milk powders: 1. Development ofsurface composition during manufacture. J. Food Eng. 94,163–168.9 3 ( 2 0 1 5 ) 156–165 165
Koc, B., Yilmazer, M.S., Balkır, P., Ertekin, F.K., 2010. Spray dryingof yoghurt: optimization of process conditions for improvingviability and other quality attributes. Dry Technol. 28 (4),495–507.
Koc, M., Koc, B., Susyal, G., Yilmazer, M.S., Ertekin, F.K.,Bagdatlıoglu, N., 2011. Functional and physicochemicalproperties of whole egg powder: effect of spray dryingconditions. J. Food Sci. Technol. 48 (2), 141–149.
Kumar, P., Mishra, H.N., 2004. Yoghurt powder – a review ofprocess technology, storage and utilization. Food Bioprod.Process. 82 (C2), 133–142.
Kurozawa, L.E., Park, K.J., Hubinger, M.D., 2011. Spray drying ofchicken meat protein hydrolysate: influence of processconditions on powder property and dryer performance. DryTechnol. 29 (2), 163–173.
Mujumdar, A.S., Law, C.L., 2010. Drying technology: trends andapplications in postharvest processing. Food BioprocessTechnol. 3 (6), 843–852.
Mujumdar, A.S., Huang, L.X., 2007. Global R&D needs in drying.Dry Technol. 25 (4), 647–658.
Myers, R.H., Montgomery, D.C., 2002. Response SurfaceMethodology: Process and Product Optimization usingDesigned Experiments, second ed. John Wiley and Sons, Inc.,New York.
Palombo, R., Gertler, A., Saguy, I., 1984. A simplified method fordetermination of browning in dairy powders. J. Food Sci. 49(6), 1609.
Schuck, P., 2002. Spray drying of dairy products: state of art. DairySci. Technol. 82, 375–382.
Stat-Ease, 2005. Stat-Ease Design Expert Package, Version 7.0.0.Stat-Ease, Minneapolis, USA.
Vignolles, M.L., Jeantet, R., Lopez, C., Schuck, P., 2007. Free fat,surface fat and dairy powders: interactions between processand product – a review. Dairy Sci. Technol. 87, 187–236.