8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 1/12

Review

Experimental design of supercritical fluid extraction – A review

K.M. Sharif a, M.M. Rahman a, J. Azmir a, A. Mohamed b, M.H.A. Jahurul c, F. Sahena c, I.S.M. Zaidul a,⇑

a Faculty of Pharmacy, International Islamic University Malaysia, Kuantan Campus, 25200 Kuantan, Pahang, Malaysiab Faculty of Pharmacy, Cyberjaya University College of Medical Sciences, 63000 Cyberjaya, Malaysiac School of Industrial Technology, Universiti Sains Malaysia, Minden, Penang 11800, Malaysia

a r t i c l e i n f o

 Article history:

Received 18 May 2013Received in revised form 16 September2013Accepted 1 October 2013Available online 11 October 2013

Keywords:

Supercritical fluid extractionExperimental designScreening designOptimization design

a b s t r a c t

Supercritical fluid extraction (SFE), a sustainable green technology leads a wide range of applicationssince the past decade. Like many other processes, SFE is sometimes criticized for its large number of fac-tors which need to be properly adjusted before every single run. Experimental design and proper statis-tical analysis with small number of trials in adjusting the SFE parameters become popular in this regard.This paper is aimed to review the common experimental designs that are frequently used in the SFE pro-cess. Utilizations of different experimental designs in SFE with the intention of either screening the mostinfluential factors or optimizing the selected factors are briefly reviewed. Strategies and recommendationaddressing the choice of appropriate design, constructing design matrix, experimental trial and data anal-ysis are discussed in this paper. For more application oriented readers of SFE, an effective and easy charton choosing proper experimental design and a list of experimental design software are also included.

 2013 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1052. Screening design for SFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

2.1. Factorial design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1072.1.1. Full factorial design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1072.1.2. Factorial factorial design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

2.2. Plackett–Burman design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1083. Optimization design for SFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

3.1. Taguchi design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1083.2. Central composite design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1093.3. Box–Behnken design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1123.4. Other optimization designs for SFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

4. Recommended Strategies of using experimental design in SFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.1. Stage 1: Appropriate experimental design selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.2. Stage 2: Experimental designing software and preparation of design matrix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1134.3. Stage 3: Experimental trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1144.4. Stage 4: Data analysis and interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

5. Concluding remark. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

1. Introduction

Experimental designs are being frequently used for the optimi-zation of different operating conditions of various processes and

for improving the chromatographic separation performance, aswell as achieving high extraction efficiency (Hibbert, 2012; Dejae-gher and Vander Heyden, 2011; Hanrahan and Lu, 2006; Ferreiraet al., 2007a). Theoretically, a number of factors have simultaneouseffect on a process. However, application of experimental design isthe most effective way to identify and optimize the significant fac-tors, and to achieve a competent result by few experimental trials.

0260-8774/$ - see front matter  2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.jfoodeng.2013.10.003

⇑ Corresponding author. Tel.: +60 9 570 4841; fax: +60 9 571 6775.E-mail address:  zaidul@iium.edu.my (I.S.M. Zaidul).

 Journal of Food Engineering 124 (2014) 105–116

Contents lists available at   ScienceDirect

 Journal of Food Engineering

j o u r n a l h o m e p a g e :  w w w . e l s e v i e r . c o m / l o c a t e / j f o o d e n g

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 2/12

Therefore, the experimental design can be defined as an approachto solve the problem systematically and, it is applied to collect dataand to analyze data for obtaining information-rich result (Gooding,2004). Optimum and valid results with a minimum effort, time andresources are the primary objectives of applying the experimentaldesign in analytical process (Montgomery, 2004; Cornell, 2011;Myers and Montgomery, 2002). In an experimental design, investi-gators deliberately maneuver one or several predetermined factorsto know their impact on experimental outcome.

Supercritical fluid extraction (SFE) is based on the solvatingproperties of supercritical fluid (SF), which can be obtained byemploying pressure and temperature above the critical point of acompound, mixture or element. Extraction by SF depends on someintrinsic tunable natures of supercritical fluid like temperature,pressure and some extrinsic features like the characteristics of the sample matrix, interaction with targeted analysts and manyenvironmental factors (Cavalcanti and Meireles, 2012; Pereiraand Meireles, 2010). A single SFE condition cannot generate en-ough information addressing all the affecting factors of SFE pro-cess. To overcome this difficulty, a large number of variablesneed to be carefully identified and investigated (Diaz and Brignole,2009; Espinosa et al., 2000; Diaz et al., 2000). By proper controllingof SFE parameters, the extractability of supercritical fluid can also

be modified which enable this process to find its field from food topesticide researches (Azmir et al., 2013; Khosravi-Darani, 2010;Brunner, 2010; Herrero et al., 2010). Moreover, a higher degree

of freedom can be obtained in extraction by SFE than the conven-tional methods, which means the number of tunable propertiesgoes higher in SFE. Thus, the tunable properties of SFE make thisprocess more unique, sensitive and specific in compared with con-ventional extraction methods.   Table 1   shows a comparison of adjusting parameters of SFE with traditional Soxhlet extractionprocess.

SFE is regarded as a green process because it does not use chem-ical solvents with drastic environmental impacts. Some applica-tions of SFE have already been commercialized and some areemerging (Machida et al., 2011). But still now it is considered asa ‘‘black box design’’ of process, because of the complex interactionof affecting factors and lack of knowledge on the in-depth fluiddynamics of supercritical fluid in extraction (Wang et al., 2010).Simple approximations of experimental units are possible to con-struct in this ‘‘black box design’’, but detail point-to-point processand extraction principle are beyond measurable. Thus, experimen-tal designs are applicable to consider many influential variablesand to generate reasonable result of interest without consideringthe unknown principle of SFE process. A graphical illustration of the position of SFE regarding numerous influential factors andthe estimation of outcome by experimental design is shown inFig. 1. According to Fig. 1, some controlling variables i.e.  x1,  x2, x3, x4; some response variables i.e.  y1,  y2,  y3,  y4 and some noisy vari-ables i.e. z 1, z 2,  z 3, z 4 co-exist in a black box process. The objectiveof experimental designs is to optimize the response variables of acertain sample by systematic modification of controlling variables.The design does not consider the variability in the responses due touncontrolled noisy variables in the experiment. The choice of experimental design for SFE depends on the objectives of the study,investigators’ intention, feasibility of experiment, cost-effective-ness, time consumption and many other important factors. Forexample, to findthe most potent factors in a particular experiment,two level factorial designs can be the choice whereas to optimizethe previously found influential factors within a predeterminedrange, more complex designs like Central Composite or Box–

Behnken design are appropriate. Based on the objectives of anexperiment, all of those designs can be categorized into two broadcategories: Screening design and optimization design. Most

 Table 1

Changeable parameters of SFE in compared with Soxhlet extraction method (adapted

from Caude and Thiébaut, 1999).

SFE Soxhlet extractionSolvent choice Solvent choiceTemperature Particle size of samplePressure Extraction timeExtraction timeSolvent flow rateSample sizeExtraction timeUse of modifier

Fig. 1.   ‘‘Black Box Design’’ of SFE (adapted from Del Castillo, 2007).

106   K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116 

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 3/12

commonly used experimental designs in SFE are Full Factorial de-sign, Fractional Factorial design, Plackett–Burman design asscreening design and Taguchi design, Central Composite designand Box–Behnken design as optimization design. Choosing theconvenient design, proper experimental design software, experi-mental trials, data analysis and interpretation are common stagesin every experimental design. These four stages are also the most

crucial to execute a fruitful use of experimental design in SFE.The basic principle of different experimental designs, their applica-tions on screening or optimizing SFE parameters are critically dis-cussed in the early stage of this review. A practical strategy of usingexperimental design in SFE and recommendation are provided inthe later stage.

2. Screening design for SFE

Although a huge number of factors influence the SFE process,some of them do not have significant effect on it. Screening themost influential factors is the primary objectives of employingexperimental design in SFE. Screening designs are used to deter-mine the most important factors and their interactions from all po-

tential factors. These kinds of designs can examine qualitative,quantitative and mixer-related factors simultaneously (Dejaegherand Vander Heyden, 2011). They are applied for improvement of separation techniques, formulations, products or processes of qual-ity control and for robustness testing. Two-level full factorial, two-level factorial factorial and Plackett-Burman design are often usedfor screening purposes (Vander Heyden et al., 2000; Dejaegher andVander Heyden, 2009; Dejaegher et al., 2009; Montgomery, 2004;Dejaegher and Vander, 2008; Lewis et al., 1999; Vander Heydenand Massart, 1996; Dejaegher and Heyden, 2007). Table 2 summa-rizes some advantages and disadvantages of different screeningdesigns.

 2.1. Factorial design

Factorial designs can study more than one factor at two or morelevels. Two-level factorial designs are used for screening purposesthat can give main and interaction effects of the considered factorswith fewer runs (Hibbert, 2012; Dejaegher and Vander Heyden,2009; Dejaegher et al., 2009; Montgomery, 2004). Various combi-nations of different levels of selected factors are generally includedin these experimental designs by which the interactions amongfactors can be assumed. Such designs are more efficient in termsof dealing with large number of variables than one variable at atime (OVAT) design. The number of factors and the number of lev-els of each factor are used to classify factorial designs. For instance,a 2 2 factorial design means it has two factors each at two levelsand a 2 3 factorial design means it has three factors and two lev-

els for each factor. Usually in factorial design, the actual variable isconverted to coded variable to give better uniformity. To deter-mine the relative effects of a factor, the coded factor level analysisis recommended because the model coefficients are dimensionless

in a coded factor level analysis and thus directly comparable(Montgomery, 2004; Kuehl and Kuehl, 2000). Analysis of variance(ANOVA) or regression analysis can be the basis for analysis of afactorial design and for model-fitting. Factorial designs are gener-ally classified into two categories; one is full factorial design andanother is factorial factorial design (Kennedy and Krouse, 1999).

 2.1.1. Full factorial designA two level full factorial design might be useful for screening

when few number of factors need to be studied (Vander Heydenet al., 2000; Dejaegher and Vander Heyden, 2009; Dejaegheret al., 2009). All possible combinations of all the input variablesand their levels are included in a two level full factorial design(Dejaegher et al., 2009; Montgomery, 2004; Dejaegher andVander, 2008; Lewis et al., 1999). Full factorial design can be de-noted by 2n when n is the number of factors (Kennedy and Krouse,1999). Each factor has two levels: ‘high’ and ‘low’ (Lundstedt et al.,1998; Deming and Morgan, 1993;   Otto, 1999; Hanrahan et al.,2005) which are expressed as ‘+1’ and ‘1’, respectively. Some-times the effect of the studied factors on a particular responsecan be described by a polynomial model and if necessary, that re-

sponse may be optimized by another design. The experimental trialnumbers are increased geometrically with the increase of factorsfor a full factorial design. For example, 22 runs of experiment arerequired to examine the effects of two factors, and 23 runs of experiment are required for three factors.

Caldera et al. (2012) optimized extraction parameters of SFE toextract antioxidant compounds (carnosol and carnosic acid) fromrosemary (Rosmarinus officinalis   L.). 23 full factorial design wasused to select important variables before optimization of the se-lected factors by Box–Behnken design. Three factors (temperature,pressure and static extraction time) were studied in this experi-ment.   Ramandi et al. (2011)   applied a full factorial design forscreening the extraction parameters of fatty acids from Borago offi-

cinalis L. flowers by SFE technique before optimization using cen-

tral composite design. Four factors: temperature, pressure,volume of modifier and static extraction time were considered asindependent variables for full factorial design. All these factorswere studied at two levels. The bioactive compounds from  Helian-

thus annuus L. (sunflower) were extracted using SFE under variousconditions (Casas et al., 2007). The impact of pre-treatment of thesample, temperature, pressure and modifiers was investigatedusing a full factorial design in this study.  Maio et al. (1997) inves-tigated the influence of different conditions of SFE on chlorinatedbenzenes and hexachlorocyclohexanes extraction from contami-nated soil by full factorial design. Extraction time, temperature,pressure and modifier concentration were considered as inputvariables. The highest important factors were temperature andtime, and the less influential factor was pressure.  Llompart et al.

(1996b) used two levels full factorial design for screening the mostinfluential variables for underivatized phenol and cresols extrac-tion from soil using SFE. Flow-rate of CO2, fluid density, extractioncell temperature, nozzle and trap temperatures, amount of 

 Table 2

Advantages and disadvantages of different screening designs.

Design Advantages Disadvantages

Two level full factorial The main effect and the interaction offactors can beidentified

The increment of number of factors leads to geometricincrement of trial number. So, the design is not feasible forscreening more than sixteen factors

Two level factorial factorial Less number of experimental trails is neededcompare to full factorial design for the equalnumber of factors

The effect from factor interactions is very limited and may bemisguided as there is no measurement of error in this design

Plackett–Burman Large number of variables can be examined with a

very few experimental trails

The design is only useful for identifying significant main

effect and do not consider any two factors interaction effect

K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116    107

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 4/12

modifier, static extraction time and the contact time between theadded modifier and sample prior to extraction were consideredfor screening test. Among these eight input variables two variables(temperature of extraction cell and density of fluid) were found toshow statistically significant impacts for all studied analytes.

 2.1.2. Factorial factorial design

To evaluate the effects of a large number of factors, many exper-imental runs are necessary which is not economically and practi-cally feasible. The effects of certain factors on a response can bestudied with a factorial factorial design under an economical andpractical manner (Luftig and Jordan, 1998). When relatively largenumber of experimental trials is needed for a full factorial design,a fraction of the full factorial design is often used to obtain thedesired information. Factorial factorial design with two levels con-tains only a fraction of full factorial design and it can examine f factors at two levels by 2 f –v experiments (v = 1,2,3, . . .,n)(Dejaegher et al., 2009; Montgomery, 2004; Dejaegher and Vander,2008; Lewis et al., 1999). A fractional factorial design is consideredas a representative subset of a full factorial design. In the initialstage of any study, it seems to be a good alternative of a fullfactorial design (Otto, 1999; Hanrahan et al., 2005).

In obtaining significant parameters of SFE for the extraction of essential oil from Diplotaenia cachrydifolia,  Khajeh (2012) appliedfactorial factorial design (252) before the employment of Box–Behnken design. The effects of five parameters, namely pres-sure, temperature, volume of modifier, dynamic extraction timesand static extraction times on the extraction yield were inspected.Extraction (static and dynamic) times showed no effect on theessential oil extraction. Khajeh (2011) tried to find out the effectsof various parameters on supercritical fluid extraction of essentialoil from Satureja hortensis. Four parameters (temperature, pressure,percentage of modifier and time of extraction) were included inthis study to determine the most significant input factors and theirinteractions by a two level factorial factorial design (24–1). Temper-ature, pressure and percentage of modifier showed significant ef-

fect but extraction time did not show significant effect on theextraction yield. Jowkarderis and Raofie (2012) used factorial fac-torial experimental design to select the important variables fromdifferent influential SFE variables for the recovery of 4-nitrotolueneand 3-nitrotoluene from soil sample. The central composite designwas used after a factorial factorial design for optimization of se-lected variables. Temperature, pressure, volume of modifier anddynamic extraction time were studied factors. The three parame-ters other than temperature were found to have statistically signif-icant effects on the extraction.

 2.2. Plackett–Burman design

Plackett–Burman design is an extensively applied design to

screen the significant variables of a process (Kennedy and Krouse,1999). It has been developed by Plackett and Burman and it is alsoa factorial factorial design with two-levels (Plackett and Burman,1946). Maximum   f  = N  1 factors with N runs can be examinedin an experiments by a Plackett–Burman design, where N is a mul-tiple of 4. If the examined factor numbers are lower than  f  = N  1,then a subset of Plackett–Burman design for N  runs may be applied(Wang and Wan, 2009). Experiments are sometimes replicated fordetermining the errors of experiment. Usually first-order polyno-mial model is fitted for the Plackett–Burman design to estimatethe effects of several factors. The significant factors of the esti-mated model can be identified by ANOVA (Plackett and Burman,1946).

Reche et al. (2002) employed Plackett–Burman design for the

evaluation of the influence of several parameters to determine  N -nitrosamines from rubbers. Pressure, temperature, extraction time

(dynamic and static), temperature of restrictor and volume of mod-ifier with two levels were the considered variables for SFE. This de-sign was carried out prior to central composite design and itrequired 12 experiments. Plackett–Burman design was used forscreening of SFE parameters before optimization of the estimatedparameters by central composite design for the phenol extractionfrom soil sample (Llompart et al., 1996a). Nine variables namely

flow rate of CO2, density of fluid, temperature of extraction cell,nozzle and trap temperature, amount of derivatizing reagent, con-centration of pyridine, static extraction time, the time of contact of the sample and derivatizing reagents prior to extraction were con-sidered in this study. Folded Plackett-Burman design with 14 de-grees of freedom was studied and it involved two centeredpoints plus 24 randomized experimental runs which mean alto-gether there were 26 runs. The applications of full factorial design,factorial factorial design and Plackett–Burman design for screeningthe most important variables of SFE are listed in Table 3.

3. Optimization design for SFE

Optimization is another practice of experimental design that

confirms the optimal conditions or settings of an experiment.The optimization approach usually starts with a screening designto select the important factors and it proceeds with an optimiza-tion design (Box et al., 2005; Bruns et al., 2006; Massart, 1997)such as Taguchi design, central composite design (CCD) or Box–Behnken design (BBD). The use of optimization designs for Super-critical fluid extraction is summarized in Table 4.

 3.1. Taguchi design

Three steps have been proposed by Taguchi to achieve a robustdesign which consists of concept design, parameter design and tol-erance design (Taguchi, 1986, 1987). Parameter design is fre-quently used for more robust or optimized output of SFE. The

stages of Taguchi design used in SFE are illustrated in Fig. 2. Tagu-chi approach uses orthogonal array for determining experimentalruns. The degrees of freedom which are necessary to study thenumber of variables and their interactions determine the size of the orthogonal array. The control factors or interactions corre-spond with the columns of orthogonal array and the carried outexperiments correspond with the rows. Orthogonal array can con-sider many factors at a time and in this design the run number in-creases with the factor numbers but much less than factorialdesign. For example, if we consider three factors with two levelsthen L 4 orthogonal array with only 4 runs is needed and twelve fac-tors with five levels can be analyzed by L 50 array, which consists 50runs. ANOVA, range analysis and analysis of signal-to-noise ratioare the main analysis methods for Taguchi experimental design.

Understanding of the process and main factors and their levels of optimal input can be analyzed by a single Taguchi method. Butthe appropriate result will mostly depend on the selection of themost influential factors and their levels which are usually fixedby the investigator.

By applying Taguchi method Chen et al. (2011) determined theoptimum conditions of SFE parameters to produce higher yield of oil from dry ginger. Dynamic extraction time, temperature, pres-sure and powder particle size were used at three levels as inputvariables for optimization. Nine experimental runs were per-formed as L 9 (34) orthogonal array. The experimental data was cal-culated by signal-to-noise ratio (S/N ratio).  Liu et al. (2009) usedorthogonal array design to optimize the extraction parametersfor seed oil extraction from   Opuntia dillenii  and to arrange the

experimental trials. The effects of three parameters on the optimalextraction yield were evaluated and a L 16 (45) orthogonal matrix

108   K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116 

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 5/12

was selected to arrange the experiments. Each of the three factorscontained four levels. Taguchi design was also applied on essentialoil extraction from clove buds (Guan et al., 2007). The effect of three parameters (temperature, pressure and particle size) wasinvestigated using three levels orthogonal array design.   Yasoubi

et al. (2007) used Taguchi experimental design to optimize the dif-ferent SFE experimental conditions in phenolic compounds extrac-tion from   Punica granatum   L. Peel. Three levels orthogonal arraydesign with a 34 matrix was used to optimize four factors namelytemperature, pressure, extraction time and percentage of modifier.Four parameters: pressure, temperature, dynamic extraction timewith ethanol as a modifier were used to optimize the SFE extrac-tion yield of tea seed oil by Taguchi experimental design (Rajaeiet al., 2005). Three levels orthogonal array (L 9) was applied in thisstudy.

 3.2. Central composite design

Box and Wilson developed Central Composite Design (CCD)

(Box and Wilson, 1951). Both linear and quadric models are al-lowed to be determined by this design. CCD seems to be a good

alternative of a three level full factorial design as it provides com-parable results with smaller number of experiments (Ferreira et al.,2007a; Tarley et al., 2009; Hibbert, 2012). CCD usually consists of afull factorial design or factorial factorial design with two levels,additional axial or star points and at least one central point of 

the experimental design (Box and Wilson, 1951; Montgomery,1997; Box et al., 2005; Bruns et al., 2006). The axial design andthe central design are almost the same for the two-level full facto-rial design except one factor that may take on levels either abovethe high level or below the low level (Kuehl and Kuehl, 2000).CCD requires experiment numbers according to   N  = k2 + 2k + c  p,

here   c  p   is replicate numbers of center point and k is factornumbers.

CCD has been applied extensively in the optimization process of an extraction step to determine the optimum conditions of differ-ent experimental parameters (Dron et al., 2002; Garrido-Lópezet al., 2006; Pizarro et al., 2006; Brachet et al., 2001; Conde et al.,2006; Gonçalves et al., 2006; Pellati et al., 2005; Baranda et al.,2005; Garcia et al., 2004; Iriarte et al., 2006; Araujo and Frøyland,

2006). Effects of three factors (pressure, temperature and time) onthe yield of phenolic and tocopherol contents of roasted wheat

 Table 3

Application of screening experimental designs for SFE.

Name of theexperimental design

Sample Extract Remark References

Full factorial design Venezuelan Rosemaryleaves

Antioxidant compounds(carnosol and carnosic acid)

Three factors (extraction pressure, extractiontemperature and static extraction time)

Caldera et al. (2012)

Full factorial design   Borago officinalis L.flower

Fatty acids Four factors (pressure, temperature, volume of modifier and static extraction time)

Ramandi et al. (2011)

Full factorial design   Helianthus annuus L. Bioactive compounds Four factors (pre-treatment of the sample,temperature, pressure and modifiers)

Casas et al. (2007)

Full factorial design Olive Leaves Waxes, hydrocarbons, squalene,b-carotene, triglycerides, a-tocopherol, b-sitosterol, andalcohols

Three factors (extraction pressure, extractiontemperature and ethanol as modifier)

Tabera et al. (2004)

Full factorial design Buttermilk Polar milk fat globule membrane(MFGM) lipids

Two factors (buttermilk source and temperature)   Astaire et al. (2003)

Full factorial design   Eucalyptus globulus

woodLipid Three factors (temperature, pressure and methanol

modifier)González-Vila et al.(2000)

Full factorial design Savory (Satureja

hortensis L.)Essential oil Two factors (temperature and pressure)   Esquıvel et al. (1999)

Full factorial design Soil Chlorinated benzenes andcyclohexane

Four factors (pressure, temperature, extractiontimeand modifier concentration)

Maio et al. (1997)

Full factorial design Soil Phenol and cresols Eight factors (carbon dioxide flow-rate, fluiddensity, extraction cell temperature, staticextraction time, nozzle and trap temperatures,

amount of methanol and the time of contactbetween the added modifier and sample prior toextraction)

Llompart et al.(1996b)

Full factorial design Liquid–solid extractioncartridge

Environmental pollutants Three factors (pressure, temperature, andextraction time)

Ho and Tang (1992)

Full factorial design Avian feed Amine hydrochloride Two factors (temperature and pressure)   Bicking (1992)Factorial factorial design Rapeseed cake Oil and minor lipid Three factors (pressure, temperature and extraction

time)Uquiche et al. (2012)

Factorial factorial design   Diplotaenia cachrydifolia   Essential oil Four factors (pressure, temperature, modifiervolume and dynamic and static extraction times)

Khajeh (2012)

Factorial factorial design   Satureja hortensis   Essential oil Four factors (pressure, temperature, percent of modifier and extraction time)

Khajeh (2011)

Factorial factorial design Soil 4-Nitrotoluene and 3-nitrotoluene

Four factors (pressure, temperature, modifiervolume, and dynamic extraction time)

 Jowkarderis andRaofie (2012)

Factorial factorial design Hantzsch reaction Formaldehyde Three factors (Pressure, temperature and, static anddynamic extraction time)

Reche et al. (2000)

Placket-Berman design Rubbers N-nitrosamines Six factors (pressure, temperature, static anddynamic time, restrictor temperature and volume

of modifier)

Reche et al. (2002)

Placket–Berman design Soil samples Phenol Nine factors (carbon d ioxide fl ow r ate, fluid d ensity,extraction cell temperature, static extraction time,nozzle and trap temperatures, amount of derivatizing reagent, pyridine concentration, andtime of contact between the derivatizing reagentsand sample prior to extraction)

Llompart et al.(1996a)

K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116    109

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 6/12

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 7/12

germ were determined during supercritical carbon dioxide extrac-tion using CCD (Gelmez et al., 2009). The dependent variables werethe yield of total phenolic compounds and the yield of totaltocopherol compounds, and their antioxidant activities.  Ghasemiet al. (2011) applied CCD to optimize the experimental parametersof SFE such as pressure, temperature, volume of modifier, extrac-tion time (static and dynamic) for essential oils extraction fromMyrtus communis L. leaves. Three level CCD was used for optimiza-tion process after screening by a two-level fractional factorial de-sign. Wang et al. (2012)  isolated essential oil from the rhizomes

of  Cyperus rotundus Linn using supercritical CO2. The four SFE fac-tors (temperature, pressure, extraction time, and flow rate of CO2)were optimized by CCD for getting optimal yield of essential oils. Asecond-order polynomial model was used to calculate the oil yield.Ramandi et al. (2011)   used SFE in different conditions for theextraction of fatty acids and essential oils from   Borago officinalis

L. flowers. CCD was used to optimize the experimental parameters

(temperature, pressure and volume of methanol as modifier) afterusing a two levels full factorial design. Brachet et al. (2000) appliedCCD as an optimization procedure for the extraction of cocaine

 Table 4 (continued)

Name of the experimentaldesign

Sample Extract Remark References

Box–Behnken design Mangosteen fruitpericarp (Garcinia

mangostana  L)

Xanthones Three factors (pressure, temperature and time)   Zarena et al.(2012)

Box–Behnken design Rapeseed Oil Three factors (pressure, temperature and extractiontime)

Cvjetko et al.(2012)

Box–Behnken design   Diplotaenia cachrydifolia   Essential oil Four factors (pressure, temperature, modifiervolume and, dynamic and static extraction times)

Khajeh (2012)

Box–Behnken design   Rosemary Leaves Antioxidant Compounds Three factors (extraction pressure, extractiontemperature and static extraction time)

Caldera et al.(2012)

Box–Behnken design   Maydis stigma   F lavonoids Three factors (temperature, pressureand co-solventamount)

Liu et al.(2011)

Box–Behnken design   Herba Moslae   Essential Oil Three factors (temperature, pressure and extractiontime)

Nie et al.(2010)

Box–Behnken design   Zingiber zerumbet  (L) Non-polar compounds Three factors (temperature, pressure and CO2

amount)Nik Norulainiet al. (2009b)

Box–Behnken design Tomato skins Lycopene Three factors (pressure, temperature and time)   Yi et al. (2009)Box–Behnken design   Pueraria lobata   Flavonoids Three factors (pressure, temperature andco-solvent

amount)Wang et al.(2008)

Box–Behnken design Grapefruit(Citrus paradisi

Macf .) seedsLimonoids and naringin Three factors (pressure, temperature and time or %

co-solvent)Yu et al. (2007)

Box–Behnken design   Spirulina platensis   Antioxidants Three factors (pressure, temperature and time)   Wang et al.(2007)

Box–Behnken design Castor oil Fatty acid composition Four factors (pressure, temperature, methanolconcentration and water concentration on the yieldof methylated castor oil)

Turner et al.(2004)

Box–Behnken design   Thymbra spicata   Aroma compounds Three factors (temperature pressure and time)   Sonsuzer et al.(2004)

Fig. 2.  Parameter design of Taguchi (adapted from Zhang et al. (2007)).

K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116    111

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 8/12

from Erythroxylum coca var.  coca leaves. After selection of signifi-cant factors (pressure, temperature, nature of the polar modifierand percentage of the polar modifier), a CCD was used to optimizethese factors to get higher extraction yield.

 3.3. Box–Behnken design

Box–Behnken design (BBD) was developed by Box and Behnken(Box and Behnken, 1960). This design consists of a factorial facto-rial design with three levels and an incomplete block design insuch a way to present as a rotatable or nearly rotatable designand to avoid the extreme vertices. BBD requires experiment num-bers based on N  = 2k(k 1) + C 0, here k is factor numbers and C 0 iscentral point numbers. BBD is useful to avoid experiments whichare in extreme conditions because the highest level and lowest le-vel combinations for every factor cannot be included in BBD.Unsatisfactory results might be avoided in BBD. The significant ef-fects of BBD on the response can be examined by ANOVA and theoptimal response can be determined by the regression model withcalculating the derivatives of the model. CCD has more factor levelsthan BBD thus BBD can be used as an economical alternative of CCD (Otto, 1999; Hanrahan et al., 2005; Ferreira et al., 2007b). Acomparison between BBD and CCD in term of experiment numbersand coefficient numbers are shown in Table 5. If an experiment has

4 factors and 16 coefficients, it will require 25 runs for both BBDand CCD but by increasing the number of factors, the run numberof BBD are slowly improved than CCD i.e. 273 runs are required for8 factors experiments by CCD whereas BBD require only 113 (seeTable 6).

Among various applications, BBD are employed to determinethe critical conditions for optimizing yield in extraction process

(Turner et al., 2004; Gfrerer and Lankmayr, 2005).   Nie et al.(2010)   employed BBD to determine the optimal conditions of supercritical fluid extraction of essential oil from  Herba Moslae.Three factors (temperature, pressure and extraction time) withthree levels of each factor were selected as input factors. The ef-fects of these factors on the extracted essential oil yield were eval-uated and total 15 runs were conducted in this study. The effects of three independent variables of SFE on the yield of lycopene extrac-tion from tomato skins using BBD were investigated by  Yi et al.(2009). Wang et al. (2008) employed SFEfor the extraction of flavo-noids from   Pueraria lobata. The optimal conditions were deter-mined using BBD by evaluating the effects of three parameters(temperature, pressure and co-solvent amount) with three levelson the flavonoid yield. Three variables (temperature, pressureand flow rate of CO2) were considered as independent variablesand total 17 experimental runs were conducted in this study.BBD was used to optimize the effects of different extraction param-eters on SFE of antioxidants extraction from   Spirulina platensis

(Wang et al., 2007). Three factors (pressure, temperature and time)were investigated at three levels in this study.   Sonsuzer et al.(2004) optimized the parameters of SFE that were affecting extrac-tion yield of aroma compounds from  Thymbra spicata  using BBD.Three independent variables at three levels were examined. Theindependent variables were temperature, pressure and time.

 3.4. Other optimization designs for SFE 

Hoogerbrugge et al. (2003)   extracted polycyclic aromatichydrocarbons from earthworms and used D-optimal experimentaldesign for optimization purpose of influential SFE parameters. Theyield of 15 polycyclic aromatic hydrocarbons (PAH) was optimized

 Table 5

Comparison of run numbers and coefficient between BBD and CCD (adapted from

Ferreira et al. (2007b)).

Factors Number of coefficient Number of experiments Efficiency

BBD CCD BBD CCD

2 6 – 9 – 0.673 10 13 15 0.77 0.674 15 25 25 0.60 0.605 21 41 43 0.61 0.496 28 61 77 0.46 0.367 36 85 143 0.42 0.258 45 113 273 0.40 0.16

 Table 6

Application of some other experimental designs for SFE optimization.

Name of experimental design Sample Extract Remarks Reference

D-optimal experimentaldesign

Earthworms Polycyclicaromatichydrocarbons

15 Polycyclic aromatic hydrocarbons (PAH) were extractedand the yield was optimized using five factors (temperature,pressure, amount of trapping sorbent, flow, and dynamicextraction time)

Hoogerbruggeet al. (2003)

Full factorial design Rapeseed cake Oil and minorlipid

Three factors (pressure, temperature and extraction time)   Uquiche et al.(2012)

Full factorial design Spearmint (Mentha spicata L.)leaves

Flavonoidcompounds

Three factors and three levels (pressure, temperature anddynamic extraction time)

Bimakr et al.(2011)

Full factorial design   Strobilanthes crispus (Pecah

Kaca) leaves

Flavonoid

compounds

Three factors at three levels (extraction pressure, extraction

temperature and dynamic extraction time)

Liza et al.

(2010)Full factorial design   Scenedesmus almeriensis   Carotenoids Two factors at five levels (pressure and temperature)   Macías-Sánchez et al.(2010)

Full factorial design   Spirulina Pacifica algae Carotenoids Four factor and three levels (temperature and pressure of thesupercritical fluid, dynamic extraction time and percentageof ethanol added as the modifier)

Careri et al.(2001)

Full factorial design Spiked soil and marinesediment standard referencematerial

Polynucleararomatichydrocarbons

Four factors and five levels(pressure, temperature, extractionfluid volume and methanol modifier concentration)

Notar andLeskovs(1997)

Genetic Algorithm (GA) Chamomile Chamomileextraction

The effect of particle diameter on extraction yield of Chamomile was investigated and the yield was optimizedusing two factors (temperature and pressure)

Rahimi et al.(2011)

Genetic Algorithm (GA) andtraditional Gradient Search(GS)

Neem seeds Nimbin GA optimization was compared with GS optimizationtechnique and GA found to be a more efficient technique.Temperature, pressure, CO2 flow rate and particle diameterwere the optimization factors

Zahedi et al.(2010)

Multivariate optimizationscheme (MOS)

Soil samples Pesticide residues MOS is highly efficient for studying a large number of variables and identifying optimal extraction conditions

Zhou et al.(1997)

112   K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116 

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 9/12

using five factors namely temperature, pressure, trapping sorbentamount, flow, and extraction time (dynamic). Liza et al. (2010) ex-tracted bioactive flavonoid from Strobilanthes crispus (Pecah Kaca)using supercritical carbon dioxide extraction. Three levels full fac-torial design was applied for optimization of input parameters forobtaining highest yield. Pressure, temperature and extraction time(dynamic) were considered as input parameters.  Macías-Sánchez

et al. (2010) observed the influence of temperature and pressureon the extraction of lutein and  b-carotene from freeze-dried pow-der of  Scenedesmus almeriensis. A three level full factorial designwas used for the optimization of two input parameters. The effectsof independent variables, temperature and pressure were opti-mized for Chamomile extraction with SFE using genetic algorithmoptimization technique (Rahimi et al., 2011). Zhou et al. (1997) ap-plied multivariate optimization scheme (MOS) to investigate theeffects of SFE parameters for the extraction of pesticide residuesfrom soil. Large number of factors can be examined and optimizedusing MOS. A list of these designs with their application on SFE isprovided in Table 5.

4. Recommended Strategies of using experimental design in SFE

4.1. Stage 1: Appropriate experimental design selection

In practice, there are two types of situations where the use of experimental design in SFE is beneficial. The first situation is tryingto understand the main controlling factors on the SFE yield and theother is measuring the optimal value for the dominant factors.Fig. 3 illustrates the approach of the selection of experimental de-sign. Before going for optimization, it is beneficial to use screeningdesign in a process which will both save time and labor. When thefactor numbers are small, the two level full factorial design can bethe choice for screening. But when the factor numbers increase(>5), Pluckett-Burman or factorial factorial become the option forscreening purpose. Minimum experimental trials are needed in aPlackett-Burman design to screen the impacting factors. The vari-able range of the selected factors may contribute to the total yieldin different ways. The best value for maximizing the yield is some-times too theoretical to achieve, hence the term optimum has beenapplied. Optimization studies try to find the most optimal levels of factors in SFE. For the optimization, choosing the levels to be inves-

tigated are crucially important. Taguchi, CCD or BBD can be usedfor optimization of small number of factors. BBD design allowsmuch economical alternative than other optimization designs.

4.2. Stage 2: Experimental designing software and preparation of 

design matrix

At this point, design matrix of an experimental study needs tobe generated. Some specialized experimental design software ex-ists commercially in market like DOE expert, Fusion Pro andModde. Some general statistical software with experimental de-sign compatibility can also be used for this purpose. Minitab, JMPand Matlab are well known package for experimental design pur-pose. A list is provided in Table 7 showing some experimental de-sign software packages. Most of the software packages usedifferent user interface to make this stage easy and convenient.Familiarity with software before data use is also recommended.Little effort is required to design and analyze data for routine anal-ysis by specialized software that is user-friendly. For example, DOEexpert has its first page where investigators have to click on appro-

priate design and select number of factors and their levels. Basedon the data this software is made to produce a design matrix forthe experiment. At this point, investigator should do trials in reallife and put their data on the previously prepared design matrix.After clicking the analysis button, investigators will find a detail re-sult of lack of fit, coefficient,   p-value and many other statistics,graphics to understand results. On the other hand, Matlab hasmore functionality on data analysis of designed experiment thanspecialized software. But for its complicated user interfaces, it isonly recommended for complex data analysis. Matlab (Mathworks,Inc., USA) has some toolbox for specialized application of experi-mental design. Neural network toolbox, genetic algorithm toolboxetc. are mainly made for more application oriented research andcan easily be incorporated for SFE optimization. In addition, De-sign-Expert and Minitab software have schemes like RSM by desir-ability function for multiple response optimization. These softwarepackages are also useful to optimize a large number of factorssimultaneously. Furthermore, neural network models based multi-ple-response optimization can also be carried out by genetic algo-rithm using the software package of Matlab.

Fig. 3.  Design selection chart.

K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116    113

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 10/12

4.3. Stage 3: Experimental trials

SFE experimental trials are sometimes troublesome because of numerous external parameters involved. Reproducibility and pre-

cision must be checked before SFE experiments are being carriedout in a particular setting. Most of the time, room temperature,water content, particle size and collection of analysts need to beconstant for complete experimental trials. Day to day variabilityand reproducibility in a particular day should be considered beforea run. Data collection process should be rigorously monitored, forexample, stopwatch and traditional clock is used in collectinginformation about time.

4.4. Stage 4: Data analysis and interpretation

The first step of data analysis should be concentrated on fittingthe value of experimental data with mathematical model applied.It is primary steps to go for further analysis and interpretation.

Sometimes investigators will find that mathematical models arenot sufficient to consider full experimental range under study. Ahigher order model can be a way to overcome this problem. Eval-uation of model fitting for experimental design in SFE study can bedone by ANOVA. A statistical significant regression and a non-sig-nificant lack of fit data represent that the experimental data arewell fitted with the model used. Another way is to look at residualvalue. If the large amount of variation can be described by residualvalues instead of regression equation, it can be interpreted as thevariation observed due to pure error will also question the modelquality (Pimentel and Barros, 1996; Cornell, 1990). A well fittedmodel should show a normal distribution of residuals if a visualgraph is created for easier inspection of model fitting. The visuali-zation of the predicted model equation can be obtained by the sur-

face response plot.

5. Concluding remark 

Experimental design is very important for conducting modernresearch using SFE to obtain precise data. Designed experimentcan give systematic investigation route and provide sequentialsteps for understanding linear, interaction and more complextypesof interaction. But successful applications of experimental designin SFE rely on both understanding about SFE and knowledge onexperimental design techniques. Some of commonly used experi-mental designs in SFE are demonstrated and the applications of dif-ferent experimental designs to screen the SFE affecting factors andconsequently to optimize of those selected factors are summarized.

This review demonstrates the dimensionality of the experimentaldesigns in SFE study.

References

Aghel, N., Yamini, Y., Hadjiakhoondi, A., Pourmortazavi, S.M., 2004. Supercriticalcarbon dioxide extraction of  Mentha pulegium  L. essential oil. Talanta 62 (2),407–411.

Akanda, M.J.H., Sarker, M.Z.I., Norulaini, N., Ferdosh, S., Khan, M.S., Azmir, J., Omar,A.M., 2012. Supercritical carbon dioxide extraction of cocoa butter analogy fatfrom mango seed kernel oil: an optimization approach of response surfacemethodology. In: Proceedings of the International Conference of ScienceTechnology and Social Science (ICSTSS 2012), Kuantan, Malaysia.

Araujo, P., Frøyland, L., 2006. Optimisation of an extraction method for thedetermination of prostaglandin E2  in plasma using experimental design andliquid chromatography tandem mass spectrometry. Journal of ChromatographyB 830 (2), 212–217.

Astaire, J., Ward, R., German, J., Jimenez-Flores, R., 2003. Concentration of polarMFGM lipids from buttermilk by microfiltration and supercritical fluidextraction. Journal of Dairy Science 86 (7), 2297–2307.

Azmir, J., Zaidul, I., Rahman, M., Sharif, K., Mohamed, A., Sahena, F., Jahurul, M.,Ghafoor, K., Norulaini, N., Omar, A., 2013. Techniques for extraction of bioactivecompounds from plant materials: a review. Journal of Food Engineering 117 (4),426–436.

Baranda, A.B., Etxebarria, N., Jiménez, R.M., Alonso, R.M., 2005. Development of aliquid–liquid extraction procedure for five 1,4-dihydropyridines calciumchannel antagonists from human plasma using experimental design. Talanta

67 (5), 933–941.Bicking, M.K., 1992. A simplified experimental design approach to optimization of 

SFE conditions for extraction of an amine hydrochloride. Journal of Chromatographic Science 30 (9), 358–360.

Bimakr, M., Rahman, R.A., Taip, F.S., Ganjloo, A., Salleh, L.M., Selamat, J., Hamid, A.,Zaidul, I., 2011. Comparison of different extractionmethods for the extractionof major bioactive flavonoid compounds from spearmint (Mentha spicata   L.)leaves. Food and Bioproducts Processing 89 (1), 67–72.

Box, G.E., Behnken, D., 1960. Some new three level designs for the study of quantitative variables. Technometrics 2 (4), 455–475.

Box, G.E., Wilson, K., 1951. On the experimental attainment of optimumconditions. Journal of the Royal Statistical Society. Series B (Methodological)13 (1), 1–45.

Box, G.E.P., Hunter, J.S., Hunter, W.G., 2005. Statistics for Experimenters: Design,Innovation, and Discovery. Wiley-Interscience, Hoboken, NJ.

Brachet, A., Christen, P., Gauvrit, J.-Y., Longeray, R., Lantéri, P., Veuthey, J.-L., 2000.Experimental design in supercritical fluid extraction of cocaine from cocaleaves. Journal of Biochemical and Biophysical Methods 43 (1), 353–366.

Brachet, A., Rudaz, S., Mateus, L., Christen, P., Veuthey, J.-L., 2001. Optimisation of accelerated solvent extraction of cocaine and benzoylecgonine from cocaleaves. Journal of Separation Science 24 (10–11), 865–873.

Brunner, G., 2010. Applications of supercritical fluids. Annual Review of Chemicaland Biomolecular Engineering 1, 321–342.

Bruns, R.E., Scarminio, I.S., Barros, N.B.d., 2006. Statistical Design – Chemometrics.Elsevier, Amsterdam.

Caldera, G., Figueroa, Y., Vargas, M., Santos, D.T., Marquina-Chidsey, G., 2012.Optimization of supercritical fluid extraction of antioxidant compoundsfrom Venezuelan rosemary leaves. International Journal of Food Engineering 8(4).

Careri, M., Furlattini,L., Mangia, A.,Musci, M.,Anklam, E., Theobald, A.,Von Holst, C.,2001. Supercritical fluid extractionfor liquid chromatographic determinationof carotenoids in Spirulina Pacifica algae: a chemometric approach. Journal of Chromatography A 912 (1), 61–71.

Casas, L., Mantell, C., Rodriguez, M., Torres, A., Macias, F., Martínez de la Ossa, E.,2007. Effect of the addition of cosolvent on the supercritical fluid extraction of bioactive compounds from   Helianthus annuus  L. The Journal of SupercriticalFluids 41 (1), 43–49.

Caude, M., Thiébaut, D., 1999. Practical Supercritical Fluid Chromatography andExtraction. Harwood Academic Publishers, Australia.

 Table 7

Software packages for experimental design and analysis (adapted from   Hibbert (2012)).

Specialized experimental design and analysis software

Design-Expert Stat-Ease Inc.,  http://www.stateease.com/   Experimental designFusion Pro S-Matrix Corporation,  http://www.smatrix.com/   Experimental designModde Umetrics,  http://www.umetrics.com/moddde   Experimental designUnscrambler Camo AS,  http://www.camo.com   Experimental designDOE++ ReliaSoft Corporation., http://www.doe.reliasoft.com/   Experimental designDOE Wisdom Launsby Consulting.  http://www.launsby.com/   Experimental designDOEpack Productivity-Quality Systems, Inc.  http://www.pqsystems.com/   Experimental design

General statistical software with experimental design compatibility

 JMP SAS Institute Inc.,  http://www.sas.com/   Statistical SoftwareMinitab Minitab Inc.,  http://www.minitab.com   Statistical SoftwareSTATISTICA StatSoft, http://www.statsoft.com   Statistical SoftwareMatlab The Mathworks Inc.,  http://www.mathworks.com.   Statistical SoftwareOrigin Microcal Software,  http://www.originlab.com/   Statistical SoftwareStatgraphics Startpoint Technologies, http://www.startgraphics.com   Statistical Software

114   K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116 

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 11/12

8/12/2019 1-s2.0-S0260877413005128-main

http://slidepdf.com/reader/full/1-s20-s0260877413005128-main 12/12

Llompart, M., Lorenzo, R., Cela, R., 1996a. Multivariate optimization of supercriticalfluid derivatization and extraction of phenol in soil samples. Journal of Chromatographic Science 34 (1), 43–51.

Llompart, M., Lorenzo, R., Cela, R., 1996b. Optimization of supercritical fluidextraction of phenol and cresols in soil samples. Journal of Chromatography A723 (1), 123–134.

Lopez-Sebastian, S., Ramos, E., Ibáñez, E., Bueno, J.M., Ballester, L., Tabera, J., Reglero,G., 1998. Dearomatization of antioxidant rosemary extracts by treatment withsupercritical carbon dioxide. Journal of Agricultural and Food Chemistry 46 (1),13–19.

Luftig, J.T., Jordan, V.S., 1998. Design of Experiments in Quality Engineering.McGraw-Hill, New York.

Lundstedt, T., Seifert, E., Abramo, L., Thelin, B.,Nyström, Å.,Pettersen, J., Bergman,R.,1998. Experimental design and optimization. Chemometrics and IntelligentLaboratory Systems 42 (1), 3–40.

Machida, H., Takesue, M., Smith, R.L., 2011. Green chemical processes withsupercritical fluids: properties, materials, separations and energy. The Journalof Supercritical Fluids 60, 2–15.

Macías-Sánchez, M., Fernandez-Sevilla, J., Fernández, F., García, M., Grima, E.M.,2010. Supercritical fluid extractionof carotenoids from Scenedesmus almeriensis.Food Chemistry 123 (3), 928–935.

Maio, G., von Holst, C., Wenclawiak, B.W., Darskus, R., 1997. Supercritical fluidextraction of some chlorinated benzenes and cyclohexanes from soil:optimization with fractional factorial design and simplex. AnalyticalChemistry 69 (4), 601–606.

Majdi, S., Barzegar, M., Jabbari, A., AghaAlikhani, M., 2012. Supercritical fluidextraction of tobacco seed oil and its comparison with solvent extractionmethods. Journal of Agricultural Science and Technology 14 (5), 1053–1065.

Massart, D.L., 1997. Handbook of Chemometrics and Qualimetrics Part A. Elsevier,Amsterdam.Montgomery, D.C., 2004. Design and Analysis of Experiments. Wiley, New York.Myers, R.H., Montgomery, D.C., 2002. Response Surface Methodology: Process and

Product Optimization Using Designed Experiments, 2 ed. Wiley, New York.Nie, S.-P., Li, J.-E., Yang, C., Qiu, Z.-H., Xie, M.-Y., 2010. Optimization of supercritical

fluid extraction of essential oil from herba moslae by response surfacemethodology and its chemical composition analysis. Food Science andTechnology Research 16 (3), 185–190.

Nik Norulaini, N.,Anuar, O.,Abbas, F.,Fatehah,M., Mohd Omar, A.,Sahena, F.,Zaidul,I., 2009a. Optimization of supercritical CO2   extraction of    Anastaticahierochuntica. Food and Bioproducts Processing 87 (2), 152–158.

Nik Norulaini, N., Anuar, O., Omar, A., AlKarkhi, A.F., Setianto, W.B., Fatehah, M.,Sahena, F., Zaidul, I., 2009b. Optimization of SC–CO2  extraction of zerumbonefrom Zingiber zerumbet  (L) Smith. Food Chemistry 114 (2), 702–705.

Notar, M., Leskovs, H., 1997. Optimisation of supercritical fluid extraction of polynuclear aromatic hydrocarbons from spiked soil and marine sedimentstandard reference material. Fresenius’ Journal of Analytical Chemistry 358 (5),623–629.

Nyam, K., Tan, C.P., Lai, O.M., Long, K., Che Man, Y., 2010. Optimization of supercritical fluid extraction of phytosterol from roselle seeds with acentral composite design model. Food and Bioproducts Processing 88 (2),239–246.

Otto, M., 1999. Chemometrics: Statistics and Computer Application in AnalyticalChemistry. Wiley-VCH, Weinheim, New York.

Pellati, F., Benvenuti, S., Yoshizaki, F., Bertelli, D., Rossi, M.C., 2005. Headspace solid-phase microextraction-gas chromatography–mass spectrometry analysis of thevolatile compounds of Evodia species fruits. Journal of Chromatography A 1087(1–2), 265–273.

Peng, J.,Dong, F., Xu, Q.,Xu, Y., Qi, Y., Han, X.,Xu, L., Fan, G., Liu, K., 2006. Orthogonaltest design for optimization of supercritical fluid extraction of daphnoretin, 7-methoxy-daphnoretin and 1, 5-diphenyl-1-pentanone from   Stellerachamaejasme   L. and subsequent isolation by high-speed counter-currentchromatography. Journal of Chromatography A 1135 (2), 151–157.

Pereira, C., Meireles, M.A., 2010. Supercritical fluid extraction of bioactivecompounds: fundamentals, applications and economic perspectives. Food andBioprocess Technology 3 (3), 340–372.

Pimentel, M.F., Barros, N., 1996. Calibração: Uma revisão para químicos analíticos.

Química Nova 19 (3), 268–277.Pizarro, C., González-Sáiz, J.M., Pérez-del-Notario, N., 2006. Multiple response

optimisation based on desirability functions of a microwave-assisted extractionmethod for the simultaneous determination of chloroanisoles andchlorophenols in oak barrel sawdust. Journal of Chromatography A 1132 (1–2), 8–14.

Plackett, R.L., Burman, J.P., 1946. The designof optimum multifactorial experiments.Biometrika 33 (4), 305–325.

Rahimi, E., Prado, J., Zahedi, G., Meireles, M., 2011. Chamomile extraction withsupercritical carbon dioxide: mathematical modeling and optimization. The Journal of Supercritical Fluids 56 (1), 80–88.

Rajaei, A., Barzegar, M., Yamini, Y., 2005. Supercritical fluid extraction of tea seed oiland its comparison with solvent extraction. European Food Research andTechnology 220 (3–4), 401–405.

Ramandi, N.F., Najafi, N.M., Raofie, F., Ghasemi,E., 2011. Central compositedesign forthe optimization of supercritical carbon dioxide fluid extraction of fatty acidsfrom Borago officinalis L. flower. Journal of Food Science 76 (9), C1262–C1266.

Reche, F., Garrigós, M., Sánchez, A., Jiménez, A., 2000. Simultaneous supercritical

fluid derivatization and extraction of formaldehyde by the Hantzsch reaction. Journal of Chromatography A 896 (1), 51–59.

Reche, F., Garrigós, M., Marın, M., Cantó, A., Jimenez, A., 2002. Optimization of parameters for the supercritical fluid extraction in the determination of N-nitrosamines in rubbers. Journal of Chromatography A 963 (1), 419–426 .

Sarker, M.Z.I., Selamat, J., Habib, A.S.M.A., Ferdosh, S., Akanda, M.J.H., Jaffri, J.M.,2012. Optimization of supercritical CO2  extraction of fish oil from viscera of African catfish (Clarias gariepinus). International Journal of Molecular Sciences13 (9), 11312–11322.

Sonsuzer, S., Sahin, S., Yilmaz, L., 2004. Optimization of supercritical CO2 extractionof  Thymbra spicata oil. The Journal of Supercritical Fluids 30 (2), 189–199 .

Sun, Y., Wei, L., Wang, J., Bi, J., Liu, Z., Wang, Y., Guo, Z., 2010. Optimization of 

supercritical fluid extraction of saikosaponins from   Bupleurum falcatum  withorthogonal array design. Journal of Separation Science 33 (8), 1161–1166.

Tabera, J., Guinda, Á., Ruiz-Rodríguez, A., Señoráns, F.J., Ibáñez, E., Albi, T., Reglero,G., 2004. Countercurrent supercritical fluid extraction and fractionation of high-added-value compounds from a hexane extract of olive leaves. Journal of Agricultural and Food Chemistry 52 (15), 4774–4779.

Taguchi, G., 1986. Introduction to Quality Engineering: Designing Quality intoProducts and Processes. Asian Productivity Organization, Tokyo, Japan.

Taguchi, G., 1987. System of Experimental Design: Engineering Methods toOptimize Quality and Minimize Costs. UNIPUS; American Supplier Institute,White Plains; Dearborn.

Tam, C., Yang, F., Zhang, Q., Guan, J., Li, S., 2007. Optimization and comparison of three methods for extraction of volatile compounds from   Cyperus rotundusevaluated by gas chromatography–mass spectrometry. Journal of Pharmaceutical and Biomedical Analysis 44 (2), 444–449.

Tarley, C.R.T., Silveira, G., dos Santos, W.N.L., Matos, G.D., da Silva, E.G.P., Bezerra,M.A., Miró, M., Ferreira, S.L.C., 2009. Chemometric tools in electroanalyticalchemistry: methods for optimization based on factorial design and response

surface methodology. Microchemical Journal 92 (1), 58–67.Turner, C., Whitehand, L.C., Nguyen, T., McKeon, T., 2004. Optimization of asupercritical fluid extraction/reaction methodology for the analysis of castor oilusing experimental design. Journal of Agricultural and Food Chemistry 52 (1),26–32.

Uquiche, E., Romero, V., Ortíz, J., del Valle, J., 2012. Extraction of oil and minor lipidsfrom cold-press rapeseed cake with supercritical CO2. Brazilian Journal of Chemical Engineering 29 (3), 585–598.

Vander Heyden,Y., Massart, D.,1996.Review of theuse of robustness andruggednessin analytical chemistry. In: Hendriks, M.W.B.d.B.J.H., Smildw, A.K. (Eds.), DataHandling in Science and Technology. Elsevier, Amsterdam, pp. 79–147.

Vander Heyden, Y., Perrin, C., Massart, D., 2000. Optimization strategies for HPLCand CZE. Handbook of Analytical Separations 1, 163–212.

Wang, J., Wan, W., 2009. Experimental design methods for fermentative hydrogenproduction: a review. International Journal of Hydrogen Energy 34 (1), 235–244.

Wang, L., Pan, B., Sheng, J., Xu, J., Hu, Q., 2007. Antioxidant activity of  Spirulina platensis  extracts by supercritical carbon dioxide extraction. Food Chemistry105 (1), 36–41.

Wang, L., Yang, B., Du, X., Yi, C., 2008. Optimisation of supercritical fluid extractionof flavonoids from Pueraria lobata. Food Chemistry 108 (2), 737–741.Wang, J., Sun, Z., Dai, Y., Ma, S., 2010. Parametric optimization design for

supercritical CO2  power cycle using genetic algorithm and artificial neuralnetwork. Applied Energy 87 (4), 1317–1324.

Wang,H., Liu, Y., Wei, S., Yan, Z., 2012. Applicationof response surface methodologyto optimise supercritical carbon dioxide extraction of essential oil from Cyperusrotundus  Linn. Food Chemistry 132 (1), 582–587.

Xiao, J.B., Chen, J.W., Xu, M., 2007. Supercritical fluid CO2 extraction of essential oilfrom   Marchantia convoluta: global yields and extract chemical composition.Electronic Journal of Biotechnology 10 (1), 141–148.

Yasoubi, P., Barzegar, M., Sahari, M., Azizi, M., 2007. Total phenolic contents andantioxidant activity of pomegranate (Punica granatum L.) peel extracts. Journalof Agricultural Science and Technology 9 (1), 35–42.

Yi, C., Shi, J., Xue, S.J., Jiang, Y., Li, D., 2009. Effects of supercritical fluid extractionparameters on lycopene yield and antioxidant activity. Food Chemistry 113 (4),1088–1094.

Yu, J., Dandekar, D.V., Toledo, R.T., Singh, R.K., Patil, B.S., 2007. Supercritical fluidextraction of limonoids and naringin from grapefruit (Citrus paradisi  Macf.)

seeds. Food Chemistry 105 (3), 1026–1031.Zahedi, G., Azarpour, A., 2011. Optimization of supercritical carbon dioxide

extraction of Passiflora seed oil. TheJournal of Supercritical Fluids 58 (1), 40–48.Zahedi, G., Elkamel, A., Lohi, A., 2010. Genetic algorithm optimization of 

supercritical fluid extraction of nimbin from neem seeds. Journal of FoodEngineering 97 (2), 127–134.

Zarena, A., Sachindra, N., Udaya Sankar, K., 2012. Optimisation of ethanol modifiedsupercritical carbon dioxide on the extract yield and antioxidant activity fromGarcinia mangostana L. Food Chemistry 130 (1), 203–208.

Zhang, J.Z., Chen, J.C., Kirby, E.D., 2007. Surface roughness optimization in an end-milling operation using the Taguchi design method. Journal of MaterialsProcessing Technology 184 (1–3), 233–239.

Zhou, M., Trubey, R.K., Keil, Z.O., Sparks, D.L., 1997. Study of the effects of environmental variables and supercritical fluid extraction parameters on theextractability of pesticide residues from soils using a multivariate optimizationscheme. Environmental Science & Technology 31 (7), 1934–1939.

Zhou, J., Ma, X.m., Qiu, B.H., Chen, J.x., Bian, L., Pan, L.m., 2013. Parametersoptimization of supercritical fluid-CO2 extracts of frankincense using response

surface methodology and its pharmacodynamics effects. Journal of SeparationScience 36 (2), 383–390.

116   K.M. Sharif et al. / Journal of Food Engineering 124 (2014) 105–116