-
1 23
Food and Bioprocess TechnologyAn International Journal ISSN
1935-5130 Food Bioprocess TechnolDOI 10.1007/s11947-013-1136-2
Optimization, Modeling, and OnlineMonitoring of the Enzymatic
Extraction ofBanana Juice
Vrani Ibarra-Junquera, Pilar Escalante-Minakata, Arturo Moiss
Chvez-Rodrguez, Isabel Alicia Comparan-Dueas, et al.
-
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media New York. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com.
-
ORIGINAL PAPER
Optimization, Modeling, and Online Monitoringof the Enzymatic
Extraction of Banana Juice
Vrani Ibarra-Junquera & Pilar Escalante-Minakata &
Arturo Moiss Chvez-Rodrguez &Isabel Alicia Comparan-Dueas &
Juan Alberto Osuna-Castro & Jos de Jess Ornelas-Paz & Jaime
David Prez-Martnez & Cristbal No Aguilar
Received: 25 May 2012 /Accepted: 20 May 2013# Springer
Science+Business Media New York 2013
Abstract This article focuses on the optimization, model-ing,
and online monitoring of banana juice productionthrough an
enzymatic method. In order to perform thistask, a batch reactor was
designed with automatic controlover the temperature and the
agitation speed as well asonline monitoring of torque. The
experiments were carriedout with the Musa AAA Cavendish banana
variety (Enanogigante), the main variety planted in Mexico. Three
differentripening stages were evaluated. Optimization of juice
extrac-tion was evaluated as a function of the pulp/water
relationshipand the concentration of the enzyme complex. The
resultsshowed that the adding of water had no influence on
theextraction of banana juice, and the optimal enzyme
concen-tration per kilogram of banana pulp was found. Based on
afuzzy logic approach, it was possible to relate the initial
torque
with the ripeness stage. Furthermore, an observable
dynamicalmodel based on ordinary differential equations and fuzzy
logicis presented. With this model, the relationship between
thetorque dynamic and the instant juice yield was found to dependon
the amount of enzyme, the temperature, and the maturitystage of the
banana used. In addition, a principal componentsanalysis was used
to classify and to relate the final juicecharacteristics (e.g., L,
a, and b colorimetric components) tothe processing conditions and
the final appreciation of a groupof sensorial panelists.
Additionally, a robust observer wasdesigned and implemented to
filter the noise present in thetorque signal and to predict the
instant juice yield.
Keywords Banana . Juice . Modeling . Observer .
Optimization . Processing
Introduction
Bananas are one of the worlds most important food crops,consumed
by millions as part of a daily diet and for nutrientenrichment
(Mohapatra et al. 2011) since they are an importantsource of
polyphenols,minerals, and carbohydrates (Kiyoshi andWahachiro 2003;
Wall 2006; Escalante-Minakata et al. 2013).Furthermore, the banana
is widely appreciated for its flavor andaroma (Boudhrioua et al.
2003; Mohapatra et al. 2011). Thoughcommonly consumed as fresh
fruit, bananas have an unfortu-nately short shelf-life due to
softening and thus are often used inbanana juice processing as an
alternative (Kyamuhangire et al.2002; Lee et al. 2006a, b;
Lpez-Nicols et al. 2007; Mohapatraet al. 2011; Chvez-Rodrguez et
al. 2013).
A relevant issue in banana juice extraction is the retentionof
the juice in the pulp due to the great amount of polysac-charides,
which prevent the release of intra-cell components,affecting the
extraction yield and the clarification. The juicecan be extracted
by a mechanical press and/or through the
V. Ibarra-Junquera (*) : P. Escalante-Minakata :A. M.
Chvez-Rodrguez : I. A. Comparan-DueasBioengineering Laboratory,
University of Colima, CarreteraColima-Coquimatln, Km 9,Coquimatln,
Colima State, Mexicoe-mail: [email protected]
J. J. Ornelas-PazCentro de Investigacin en Alimentacin y
Desarrollo A.C.(CIAD), Unidad Cuauhtmoc, Chihuahua State,
Mexico
J. D. Prez-MartnezFaculty of Chemical Sciences,
AutonomousUniversity of San Luis Potos,San Luis Potos, S.L.P.
State, Mexico
J. A. Osuna-CastroFaculty of Biological and Agricultural
Sciences,University of Colima, km 40 Autopista
Colima-Manzanillo,28100, Tecomn, Colima, Mexico
C. N. AguilarDepartment of Food Science and Technology, School
ofChemistry, Autonomous University of Coahuila, SaltilloCoahuila
State, Mexico
Food Bioprocess TechnolDOI 10.1007/s11947-013-1136-2
Author's personal copy
-
action of pectinolytic enzymes, such as pectinase
andpolygalactouronase (Casimir and Jayaraman 1971; Viquez etal.
1981; Kyamuhangire et al. 2002; Lee et al. 2006a). For thatreason,
banana juice extraction can be considered a bioprocesswhen an
enzymatic bioreactor is at its core. Such a processdemands high
levels of quality and safety in food produc-tion, which calls for
high standards in quality assurance andprocess control; satisfying
this demand, in turn, requiresappropriate analytical tools for food
analysis both duringand after production (Zhong 2010). Desirable
features ofsuch tools include speed, ease of use, minimal or
nosample preparation, and the avoidance of sample destruc-tion. It
is also necessary to control and optimize the biore-actor
environment via operating variables in order to favorthe desired
functions and achieve cost-effective, large-scalemanufacture (Zhong
2010). Traditionally, a method toachieve this goal utilizes
laboratory instruments to treatsamples taken from the processing
plant and analyzed off-line (Ibarra-Junquera et al. 2010). This
allows processinvestigation, checking of the specifications of raw
mate-rials and final products, which aids in management anddecision
making (Cozzolino et al. 2011). However, recentdevelopments in
on-line measurements offer the possibilityof following process
dynamics, achieving better and morerapid control, and eliminating
questions about the statisticalsignificance of quality control
based on sampling (Zhang2009; Ibarra-Junquera et al. 2010;
Cozzolino et al. 2011).Several studies have described on-line
instrumentation fordetermining the characteristics of particle
systems in situ inthe processing plant and that describe how
on-line mea-surements from such instruments lead to process
improve-ments. In particular, better monitoring and control of
bio-reactors requires reliable on-line estimation of process
vari-ables and parameters that often cannot be measured
directly(Zhang 2009; Ibarra-Junquera et al. 2010).
Therefore, this study has three main goals. The first goalis to
optimize the enzymatic method for banana juice ex-traction by
finding the minimum amount of water andenzymatic complex needed to
obtain the maximum juiceyield in a given span of time. Second, this
study aims todevelop a minimalist model to describe, in general
terms,banana juice extraction dynamics, allowing its usage for
on-line monitoring and automatic control. Thus, such a modelshould
include on-line measurable variables and the juiceyield as one of
its states, and then construct a model-basedobserver algorithm to
infer in real time the juice yield fromthe on-line measurable
output. Under the hypothesis that thetorque evolution can be
associated with the juice yielddynamic, a batch reactor was
designed with automatic con-trol over the temperature and the
agitation speed and withon-line monitoring of torque. The studys
third goal is torelate the visual characteristics of the final
banana juice tothe processing conditions.
Methodology
Bioreactor
The enzymatic juice extractions were performed in a 4-Lstainless
steel batch-jacketed reactor (i.d. 12.5 cm),equipped with automatic
control over the temperature, basedon a cooling thermostat with a
precision of 0.02 C (RE630S; Lauda Eco Silver, Germany). The
bioreactor agitationspeed control is composed of the Compact
cRio-9074 dataacquisition system (DAS) (National Instruments, TX,
USA),a PC for the control of the DAS, a stepper motor (NEMA
23Stepper Motor; National Instruments) connected to aMicrostepping
Drive (National Instruments), and two inde-pendent power supplies.
The stepper motor movement isdirected by using an application
developed within theLabview platform (National Instruments). An
open-paddleimpeller (width 9.6 cm, height 12.3 cm) is connected to
themotor through a drill chuck. The torque was monitoredonline
using a rotary torque sensor with a maximum capac-ity of 2 Nm
(FSH01979; FUTEK, CA, USA) connected to aPC through a data
acquisition system from National Instru-ments (Cryo-9074) and
Labview-based interface.
Biological Material
Bananas of the cultivar Enano gigante (Musa AAA, sub-group
Cavendish) were acquired in the local market whenthey were green
(Fig. 1) with a minimum length of 13 cm.The banana bunches were
ripened in laboratory conditionsat 20 C without atmospheric control
for different storagetimes. The color of the peel was monitored by
comparisonwith a scale ranging from 1 to 3, with values
representingthe following stages of ripeness: Green (1), ranging
frommore yellow than green to yellow with green endings;Yellow (2),
ranging from completely yellow to yellow withslight brown specks;
and Brown (3), which included yellowwith many brown specks.
Additionally, the parts of individ-ual samples were weighed in an
analytical balance to deter-mine the pulp-to-peel ratio. The
results were expressed asthe percent weight of pulp relative to
peel.
Fig. 1 Peel color changes in different stages of banana
ripeness, thenumbers 1, 2, and 3 corresponds to green, yellow, and
brown,respectively
Food Bioprocess Technol
Author's personal copy
-
Juice Extraction Procedure
The banana fruits were washed in 0.2 % (v/v) sodiumhypochlorite
aqueous solutions for 5 min, peeled, and cutinto pieces. In order
to homogenize the banana pulp, appro-priate volumes of water were
added to achieve a bananapulp/water ratio of 0.8, 0.9, and 1.0.
Then, the mixture washomogenized to puree in an electrical blender.
The enzy-matic extractions were performed using a commercial
mix-ture of three plant-cell-wall-degrading
enzymespectinase,cellulase, and hemicellulase (Macerex PM, ENMEX,
S.A.de C.V. (2003)), which hydrolyze the polysaccharide sub-strates
from banana cell wallranging from 100 to 600 Lof enzyme mixture per
kilogram of puree. The enzymaticextractions took place in a batch
reactor (filled with 3 kg ofbanana puree) and continuously agitated
(100 rpm) at roomtemperature (25 C) for 2 h. Subsequently, the
juice wascentrifuged at 12,000g using a centrifuge (RC6+;
Sorvall,Newtown, CT, USA) in an SLA-3000 rotor for 15 min, andthe
supernatant was collected and filtered. Finally, the totalsoluble
solids content was determined for each banana juiceobtained, using
a digital refractometer (RP-101; Atago, To-kyo, Japan) with a scale
ranging between 0 and 45 degreesBrix (Bx). The soluble solids
content were reported asdegrees Brix.
Juice Yield
The banana juice yield was calculated based on the weightof the
resulting juice, the added water, and the banana pulp,according to
the following equation:
Juice yield % jw g ww g bw g 100 1
where jw is the amount of juice recovered, ww is the weightof
the added water, and bw is the weight of the total bananapuree
used. All weight measurements were expressed ingrams (g). The juice
was recovered by centrifugation atthe same conditions mentioned in
the juice extraction meth-odology. The samples weights were
obtained through ananalytical balance (OHAUS; Explorer Pro, Pine
Brook,NJ, USA; accuracy 0.1 mg).
Protein Quantification
The protein concentration of the enzymatic complex wasdetermined
by the Bradford method (1976) as modified byBioRad (using bovine
serum albumin protein as standard).The samples were diluted to a
ratio of 1:20 by addingdistilled water. The absorbance of the
samples was deter-mined at 595 nm using a lambda 25 UVVis
spectropho-tometer (Perkin-Elmer Instrument, USA).
Colorimetric Analysis
Analysis of color variation was performed with a totalvolume of
20 mL from each banana juice sample anddispensed into separate
Petri dishes. The color values ofeach sample were obtained with a
LABSCAN XE chromameter (Hunterlab, VA, USA) at room temperature.
Equip-ment was set up for illuminant D65 and 10 observer
angle.CIE-Lab values of L (lightness), a (redness), and
b(yellowness) were determined to describe the precise loca-tion of
a color inside a three-dimensional visible colorspace.
Statistical Analysis
All statistical analyses were carried out using MatLab soft-ware
(MathWorks Inc., USA). The analysis of variances(ANOVA) was applied
to compare the mean values ofsamples according to the different
factors. In addition, prin-cipal component analysis (PCA) was used
as a data multi-variate technique. The purpose of this method is to
decom-pose the data matrix and concentrate the source of
variabil-ity in the data into the first few principal components.
Here,PCA was applied to two different groups of values: thematrix
of maturity mean attribute ratings across bananasamples (initial
torque and the peel-to-pulp ratio) and thematrix of mean color
measurements (L, a, and b). The twocases were also analyzed by
cluster analysis (average link-age method) in an attempt to
classify the samples in terms oftheir maturity and processing
method.
Visual Evaluation
The juice data, coming from the colorimetric analysis,
wereevaluated using PCA and cluster analysis. As an
additionalreference, the resulting groups were classified in terms
ofvisual acceptance by a five-judge panel belonging to thefaculty
of Chemical Sciences and research scholars of theBioengineering
Laboratory (both of the University Colimain Mexico). This general
appearance of the juice was mea-sured subjectively on a three-point
hedonic scale: poor(dislike), good (like), and excellent (like very
much).
Optimization Procedure
The optimization methodology used in this work is based ona
three-step process: (1) generating a mesh of experimentaldata
within the design space, (2) modeling the data setsthrough
nonlinear fitting, and (3) using the fitted modeland ANOVA analysis
to obtain the optimal condition.
The dependent variables selected for these studies wereX, enzyme
concentration (L/kg), taking values in the set{0, 25, 50, 100, 150,
300, 600 L/kg}; Y, ratio of
Food Bioprocess Technol
Author's personal copy
-
water/banana pulp taking values in the set {1, 0.9, 0.8}; andZ,
response variable. Given are a total of 63 experiments, tobe
performed in triplicate. The juice yield, as evaluated inEq. 1, was
the response variable. Based on the preliminarydata, the following
generalized MichaelisMenten responsesurface (Hirst et al. 1996) is
used to model the data behav-ior:
Z X ; Y bXYc dYZ 2
Exclusively for the optimization process, the
previouslymentioned banana juice extraction methodology was
slight-ly modified since all experiments were carried out at 50
Cfor 30 min, and the yield was calculated at the end of
theprocess.
Fuzzy Logic Approach
The complexity of biological processes often renders
im-practical the development of detailed, structured
phenome-nological models. Fuzzy modeling is based on sets of
fuzzyifthen rules derived from expert domain knowledge tohandle
uncertainty. The concept of fuzzy logic provides anatural way of
dealing with problems where the absence ofwell-defined criteria
could be a source of imprecision. Sincefuzzy systems can
simultaneously handle numerical dataand linguistic knowledge, they
provide opportunities formodeling of conditions that are inherently
imprecisely de-fined, like fruit ripeness.
The quality of banana juice is defined by its phys-ical
appearance and taste, both of which are directlyrelated to its
stage of ripeness. Thus, the ripenessstage directly influences the
quality of the final juice.Therefore, the purpose of the fuzzy
modeling is torelate the banana maturity as perceived by human
eyesbased on color and general physical characteristics(e.g.,
appearance of spots) to standard measurementsthat can allow the
modeling and monitoring of thejuice extraction process.
Triangular membership functions have been used forsimilar
problems (Sinija and Mishra 2011). For example,triangle a1b1c1
represents the membership function forripeness stage 1, triangle
a2b2c2 represents the distribu-tion function for ripeness stage 2,
etc. Figure 2 representsthe triplets associated with three ripeness
stages. The secondnumber of the triplet denotes the coordinate of
the abscissaat which the value of the membership function is 1. The
firstand third numbers of the triplet designate the distance to
theleft and right, respectively, of the second number where
themembership function is 0.
Enzymatic Juice Extraction Model
The main goal of this paper is to develop a minimalist modelto
describe, in general terms, banana juice extraction dy-namics to be
used for monitoring and control. This unstruc-tured and
nonsegregated model is given by Eqs. 3 and 4.
dx1dt
m T ; x2;0
xmax T ; x2;0
x1 3
dx2dt
m T ; x2;0
y T ; x2;0 xmax T ; x2;0 x1 4
where the state variables are as follows: x1 stands for
juiceyield and x2 stand for torque. The parameters are as follows:T
stands for temperature (C) and M for the maturity stage,i.e., M
(1,2,3). The rest of the functions are given inTable 1. The
mathematical expressions given in Eqs. 3 and4 were selected since
the preliminary data showed an expo-nential growth of the juice
yield and an exponential decay ofthe torque signal as well as a
temperature and maturity stagedependence of the kinetic rate, m,
the maximum juice yield,xmax, and the torque initial condition
x2,0. The functiony(T,x2,0) allows to relate the juice yield x1(t)
with the torque
Table 1 Functions and units used in dynamical model given by
Eqs. 3and 4
Variable Dependence Units
xmax(T,x2,0) f1(T,M)
y(T,x2,0) f2(T,M) 1/N m
m(T,x2,0) maxS kmS
kmBe M X 2;0
L/kg
maxAe
1T
( M(X2,0))2+w M(X2,0)+ C
A e Y M X 2;0
Fig. 2 Memberships function with triangular membership
distribution
Food Bioprocess Technol
Author's personal copy
-
measurements x2(t). Thus, this methodology can be consid-ered as
a data-driven approach to the modeling of theenzymatic extraction
of banana juice in which its structureand the parameter
identification is based on off-line and on-line time series
obtained from a rigorous experimental set.
Software Sensor Design
Taking into account the fact that rarely can one have asensor on
every state variable, and some form of reconstruc-tion from the
available measured output data is needed,software can be
constructed using the mathematical modelof the process to obtain an
estimate of the true state x. Eversince the original work by
Luenberger (1971), the use ofstate observers has proven useful in
process monitoring andfor many other tasks. In the sense of control
theory, analgorithm capable of giving a reasonable estimation of
theunmeasured variables of a process will be called anobserver.
Numerous attempts have been made to develop observerdesign
methods for bioprocess applications (Escalante-Minakata et al.
2009; Zhang 2009; Fernndez-Fernndezand Prez-Correa 2010). The first
systematic approach forthe development of a theory of observers was
proposedsome time ago by Krener and Isidori (1983). Nevertheless,it
is well known that classical proportional observers tend toamplify
the noise of on-line measurements, which can leadto the degradation
of the observer performance. In order toavoid this drawback, in
this paper, the observer algorithm isbased on the work of
Ibarra-Junquera et al. (2005) becausethe proposed integral observer
provides robustness againstnoisy measurements and
uncertainties.
In this section, the design of the software sensor ispresented
in which xj, for j(1n), is the naturally mea-sured state (i.e., the
variable easiest to measure). Therefore,it seems logical to take xj
as the output of the system, y=h(x)=xj. Now, considering that the
output function h(x) is con-taminated with a Gaussian noise, the
model given by theaforementioned Eqs. 3 and 4 acquires the
form:
X f X y cX
where f(X) is a matrix containing the left-hand side of Eqs.
3and 4, y represents the noisy on-line measurable output, isthe
additive bounded measurement noise, X=[x1,x2] is thevector of
states, and the vector cX=[0,1]T[x1,x2] defines theon-line
measurable output, x2 in this particular case. Then,the task of
designing an observer for the system Eqs. 3 and 4is to estimate the
vector of states X, despite the noise butconsidering that y is
measured on-line and that the system isobservable.
Results and Discussion
Optimization Procedure
Many recent works show that the use of
polysaccharidasesfacilitates the release of juice and increases the
extractionyield (Kyamuhangire et al. 2002; Sreenath et al.
1994;Cheirsilp and Umsakul 2008; Buenrostro-Figueroa et al.2010;
Bahramian et al. 2011). In addition, in the case ofbanana and mango
juice extraction, many protocols includedilution of pulp in water,
ranging from undiluted extractionprocesses (Khalil et al. 1989;
Reddy and Reddy 2005;Buenrostro-Figueroa et al. 2010) to water/pulp
ratios of1:1 (Cheirsilp and Umsakul 2008; Kyamuhangire andPehrson
1999; Onwuka and Awam 2001), 1:2 (Lee et al.2006a, b; Lee et al.
2007), 1:4 (Falade and Babalola 2004),and 1:5 (Akubor 1996).
However, none of these workspresent a justification for the
selected dilution ratio. Whilethere is clear evidence that the
application of enzyme com-plexes in the extraction of banana juice
is efficient, no studyhas reported on the variables of ripeness
stage andwater/pulp ratio.
The results of the Bradford assay show that the commer-cial
enzymatic complex contains 12.06 1.20 mgprotein/mL. The juice yield
was affected by the enzymaticcomplex concentration and water/pulp
ratio within the stud-ied conditions (P
-
develop a simplified model that relates on-line
measurablevariables with off-line ones. In addition, the resulting
modelshould allow the construction of an observer scheme to
re-build, in real time, the non-measurable on-line variables.
It has been previously reported that the use ofpolysaccharidases
improves the juice yield and modifiesthe physicochemical and
rheological properties of the mix-ture during the process (Aguilar
et al. 2008; Bahramian et al.2011). Formerly, off-line viscosity
measurements of pectinsolut ion were used to determine the activity
ofpolysaccharidases because a decrement of this parameter isrelated
to this endohydrolase activity (Combo et al. 2012).Since torque is
the measurement of the force that rotates anobject around its axis
(e.g., an open-paddle impeller), thisforce represents the
resistance that a fluid offers to rotationalmotion. Based on that
fact, Virgen-Ortz et al. (2012) mea-sured the pectinolytic activity
throughout the slope of the
decrement of the on-line torque in a pectin solution whenthe
pectinase was added. Although this procedure involvesthe on-line
measure of the torque, it is in fact an off-lineprocedure since the
analysis is performed after the enzymaticprocess has ended. Thus,
the relation between juice yield andthe torque on-line monitoring
is investigated here.
In order to effectively model the juice extraction process,96
experiments were carried out (only for dynamical model-ing
purposes), at various enzymatic complex concentrations,fruit
ripeness stages, and the process temperatures but withno additional
water, since the optimization has shown thatthis is not necessary.
From these experiments, the maximumjuice yield was identified as a
function of ripeness stage andtemperature. Then, for each
temperature, ripeness stage, andenzymatic complex concentration,
the right-hand side ofEq. 3 can be integrated as follows: ln(x1max
x1)=mt. Then,the juice yield rate m was determined by plotting the
naturallogarithm of maximum juice yield minus the instant
juiceyield (obtained from the off-line monitoring of the
process)versus time. The slope of the line is the kinetic rate m.
Thus,a specific value of m was obtained for each
temperature,enzymatic complex concentration, and ripeness
stage.
Next, the equation m E maxE kmE was used to model therelation
between the value of m and the enzymatic complexconcentration (E)
at each temperature and ripeness stage. Thenonlinear fitting was
performed in MatLab (R 2010b). Theobserved deviations when plotting
the predictedm values versusthe experimental data are graphically
presented in Fig. 4, where itis possible to appreciate the good
accuracy of the prediction.
It is important to note that the kinetic rate m decreases at50 C
in all of the enzymatic complex concentrations andripeness stages
studied. Thus, the data obtained at 50 Cwere not taken into account
for the temperature dependencestudies (although this phenomenon is
briefly studied at theend of this subsection).
The calculated values of km reveal no change as a functionof the
temperature, while the dependence of max on the
0200
400600
0.8
0.85
0.9
0.95
145
50
55
60
65
70
75
enzyme concentration (L/kg)
water/pulp ratio
juic
e yi
eld
(%)
Fig. 3 Graphic shows the nonlinear fitting of the experimental
data(blue circles) and the model given by Eq. (2). The model
estimatedparameters are a=49.5, b=225.4, c=100, and d=10, with
R2=0.9808
0 2 40
1
2
3
4 x 10-3
m
Concentration (L/kg)0 2 4
0
1
2
3
4 x 10-3
m
Concentration (L/kg)0 2 4
0
1
2
3
4 x 10-3
m
Concentration (L/kg)x 102 x 102 x 102
a) b) c)
Fig. 4 Plots show the nonlinear regression plots for kinetic of
juiceyield by the enzymatic complex, the lines (. . .), (__),
(-+-+), and (- - -)stand for the model prediction at 20, 30, 40,
and 50 C, respectively.
The symbols (), (*), (), and () stand for experimental data at
20, 30,40, and 50 C, respectively. Plots (a), (b), and (c)
correspond toripeness stages 1, 2, and 3, respectively
Food Bioprocess Technol
Author's personal copy
-
temperature was modeled using the following equation
Ae 1T . To perform the identification ofA and , a linear
fitting
using the natural logarithm of max versus the inverse
oftemperature was performed for each of the three ripenessstages.
Figure 5 graphically demonstrates the accuracy of theaforementioned
procedure.
Then, the dependence of km, A, and values with respectto the
ripeness stages were calculated. To perform the iden-tification of
these parameters, a linear fitting using thenatural logarithm of A
and versus the inverse of theripeness stages (expressed as 1, 2,
and 3) was performed,while a nonlinear correlation was needed for
the case of the and ripeness stage relation. Figure 5 graphically
demon-strates the accuracy of these results.
Data presented in Table 2 summarizes the identifiedmodel
parameters values and their units. It is worth men-tioning that the
mathematical model has been formulated insuch a way as to allow the
description of the juice produc-tion rate and its relation with the
torque over the work rangeof ripeness stages and temperatures.
At this point, twomore correlations were needed in order tofully
model the processthe relation xmax(T,M) and y(T,M). Inboth cases,
the relations given in Fig. 6 were found.
Since bananas are a climacteric fruit, when harvested at
thepreclimacteric matured green stage, the fruit undergoesvarious
physicochemical changes in terms of composition,color, texture,
aroma, and taste, pertaining to changes inmetabolic rates and
biochemical reactions like respiration,ripening, and senescence in
the climacteric phase (Mohapatraet al. 2011). In fact, such changes
in the physicochemicalproperties are the manifestation of various
complex biochem-ical reactions (Zolfaghari et al. 2010). Thus,
bananas can havedifferent uses in function of its maturity
stage.
In order to correlate the fruits physicochemical
charac-teristics (pulp-to-peel ratio, average of the data
correspond-ing to the first 10 min of torque values and the Brix
degree
measured in the obtained juice) with the visual bananaripeness
stage (Fig. 1), a PCA analysis was performed.The scatter plot for
the first two principal components ispresented in the Fig. 7, where
it is possible to appreciate anoverlap between the groups
corresponding to the fruit ripe-ness stages visually classified
according to Fig. 1. Similarly,the ANOVA analysis indicates that
there are no significantdifferences between the nearest neighbors
corresponding tothe mean values of first 10 min of the torque
signal of eachripeness stages and processing temperature, except in
theripeness extremes; however, it is possible to appreciate aclear
gradient of change in the torque signal as the ripenessstages
advance.
Thus, to correlate the initial torque measurements withthe
ripeness stages, a fuzzy logic approach is proposed thatmakes it
possible to correlate the visual criteria used toclassify the
bananas with the initial online torque measure-ments. The fuzzy
logic algorithm is based on the knowledgeacquired with the previous
results. To infer the maturitystage from on-line torque
measurements M(X2,0), it is firstnecessary to create the membership
functions for the torquemeasurements (input of the fuzzy logic
model) and the
Ln
max
Ln
max
Ln
max
1/T 1/T0.03 0.04 0.05 0.03 0.04 0.05 0.03 0.04 0.05
1/T
1 2 3
Ln K
m
Ripeness stage1 2 3
-7
-6.5
-6
-5.5
-6
-5.5
-5
Ln A
Ripeness stage1 2 3
-7
-6.5
-6
-5.5
-30
-20
-10
Ripeness stage
a) b) c)
d) e) f)
-7
-6.5
-6
-5.5
-1.5
-1
-0.5
Fig. 5 The plots (a), (b), and(c) graphically show
thecorrelation between the naturallogarithm of the
experimentalvalues and inverse oftemperature (C) for eachripeness
stage. From left toright, the results for ripenessstages 1, 2, and
3 are shown.Plots (d) and (e) correspond tothe correlation between
thenatural logarithm of theexperimental and values and theripeness
stage, while the plot (f)corresponds to the second-orderpolynomial
used to model therelation between and theripeness stage
Table 2 Identified model parameters and functions
Parameter Value Units
xmax(T,x2,0) 0.14T+1.5M(X2,0)+62.63
y(T,x2,0) 46T1,400M(X2,0)373.33 1/N m
0.008325768
0.3308
B 0.234898 L/kg
0.25541
9.1405 C
10.4698 C
38.422 C
Food Bioprocess Technol
Author's personal copy
-
membership functions for the maturity stage (output of thefuzzy
logic model) and name them. Each membership func-tion is described
in the form of a triplet (a, b, and c) asrepresented in Fig. 2. The
triplets corresponding to the inputand output sets as well as the
group names are shown inTable 3.
The fuzzy rules were set according to the understandingof the
behavior of the system. Since the relation betweentorque and
ripeness stage is a single-input single-outputrelation, the
following rules were used: If (input is Brown)then (output is 3);
If (input is Yellow) then (output is 2); If(input is Green) then
(output is 1).
In order to infer the juice yield initial condition fromonline
torque and temperature measurements, the member-ship functions for
the error (input of the fuzzy logic control-ler) and the gains
(output of the fuzzy logic controller) areneeded. The group names
are shown in Table 3 and anillustration of the functions is shown
in Fig. 2.
Since the relation between torque and ripeness stage is
amultiple-input single-output relation, the following ruleswere
used: If (input is Brown), then (output is 3); If (inputis Yellow),
then (output is 2); If (input is Green), then(output is 1). The
rules were stated as follows: If (Torqueis Brown) and (Temperature
is 20 C), then (Output is 60); If(Torque is Brown) and (Temperature
is 30 C), then (Outputis 60); If (Torque is Brown) and (Temperature
is 40 C),
then (Output is 65); If (Torque is Brown) and (Temperatureis 50
C), then (Output is 65); If (Torque is Yellow) and(Temperature is
20 C), then (Output is 50); If (Torque isYellow) and (Temperature
is 30 C), then (Output is 55); If(Torque is Yellow) and
(Temperature is 40 C), then (Outputis 60); If (Torque is Yellow)
and (Temperature is 50 C),then (Output is 60); If (Torque is Green)
and (Temperature is20 C), then (Output is 40); If (Torque is Green)
and (Tem-perature is 30 C), then (Output is 40); If (Torque is
Green)and (Temperature is 40 C), then (Output is 45); If (Torqueis
Green) and (Temperature is 50 C), then (Output is 45).
Remarks on Enzyme Activity and Temperature
As shown Fig. 8, the reaction rate decreases at 50 C in allthree
stages of maturity. This apparently contradicts theenzymatic
complex data sheet, regarding the optimal tem-perature. Therefore,
we sought to measure the protein con-centration by Bradford both as
banana juice in distilledwater at the same pH 5 (pH featuring
banana juice) andpH 3.5, which is recommended as optimal in the
technicalsheet. The one-way ANOVA results of the Bradfordperformed
to the protein solutions have shown a significantdecrement in the
concentration when the solution has thesame pH as the juice and 50
C. This can be due to a lack ofsolubility, and this can explain
decrement in the enzyme
2030
4050
12
3
65
70
75
Ripeness stageTemperature (C)
Xm
ax (
T,M
)
2030
4050
12
3
-10000
-5000
0
Ripeness stage
y (T
,M)
Temperature (C)
Fig. 6 From left to right, theplots correspond to the
linearcorrelation between theexperimental values obtained
atdifferent temperatures andripeness stages
-3 -2 -1 0 1 2 3-2
-1
0
1
2
First Principal Component (71.1%)Sec
ond
Prin
cipa
l Com
pone
nt (
18.2
%)Fig. 7 The plot corresponds to
the scatter plot for the first twoprincipal components of
thefruit physicochemicalcharacteristics matrix; green,blue, and
brown stand forripeness stages 1, 2, and 3,respectively. The
percentage ofthe total variance explained byeach principal
component isindicated in brackets
Food Bioprocess Technol
Author's personal copy
-
activity. It is known that the tridimensional structure of
proteinsis sensitive to minor changes in factors in the
environment,including pH, temperature, and medium composition, and
suchstructure changes could alter the catalysis activity of the
pro-teins (Srivatsa 1996).
Colorimetric Analysis
The color of a product is one of the most importantproperties
that influence the consumers response to it.Therefore, one of the
objectives of this study was todetermine how the characteristic of
the raw material andthe process conditions influence the final
color attributesof the banana juice. In order to allow the
interpretationthe color results from the L.a.b. analysis, a PCA
tech-nique plus a clustering analysis were performed. Theresults
appear in Fig. 9.
Cluster analysis groups data objects with similar
charac-teristics that were also made then different from objects
ofother group. Therefore, the three clusters formed in Fig.
9correspond to the banana juice with similar colorimetric
com-ponents. The sensorial panelists were asked to categorize
thesamples corresponding to each attribute of all samples. Fromthe
PCA score, the cluster analysis, and opinions of thepanelists, it
can be concluded that the clusters of green di-amonds obtained in
Fig. 9 are the ones with the best visualattributes. This group
corresponds to ripeness stage 1 (green)and with no influence of the
temperature or the enzymeconcentration; however, it is important to
mention that theclosest neighbors were samples of the blue star
group corre-sponding to ripeness stage 2 (yellow) and produced at
tem-peratures of 40 and 50 C.
The Software Sensor
On-line availability of the juice yield measurement is
veryimportant for the control and particularly for the process
ofsupervision and fault detection; however, there exists nodevice
that provides an on-line juice yield measurement.One way to
overcome this problem is to use softwaresensors to estimate missing
state variables on-line. As itis described in the Methodology
section, a robust observerwas constructed. In order to provide the
observer withrobust properties, the following representation of the
system(Eqs. 3 and 4) is proposed:
x0 x2 5
x1 m xmaxx1 6
x2 m
Yxmaxx1 7
y0 x0 8where x0 is the dynamical extension that allows us
tointegrate the noisy signal in order to recover a filtered
Fig. 8 The plot corresponds to the correlation between the
kinetic rateand temperature in the banana juice extraction process.
Green squares,blue triangles, and red circles correspond to
ripeness stages 1, 2, and 3,with and R2 of 0.885, 0.998, and 0.992,
respectively
Table 3 Values of the different membership functions used in
thefuzzy logic algorithm that relates the on-line torque
measurements tobanana ripeness stage, i.e., M(X2,0), as well as the
values of thedifferent membership functions used in the fuzzy logic
algorithm thatinfer the initial juice yield condition [x1,0 =
x1(0)] from the on-linemeasurements of temperature and torque
Set name a b c Set type
Torque
Brown (0.008,0) (0.0113,1) (0.0145,0) Input
Yellow (0.0135,0) (0.0149,1) (0.0205,0) Input
Green (0.0195,0) (0.024,1) (0.0285,0) Input
Ripeness stage
1 (0.5,0) (1,1) (1.5,0) Output
2 (1.4,0) (2,1) (2.45,0) Output
3 (2.4,0) (3,1) (3.6,0) Output
Temperature
20 C (14,0) (20,1) (26,0) Input
30 C (24,0) (30,1) (36,0) Input
40 C (34,0) (40,1) (46,0) Input
50 C (44,0) (50,1) (56,0) Input
Initial yield
40 (36,0) (40,1) (44,0) Output
45 (41,0) (45,1) (49,0) Output
50 (46,0) (50,1) (54,0) Output
55 (51,0) (55,1) (59,0) Output
60 (56,0) (60,1) (64,0) Output
65 (61,0) (65,1) (69,0) Output
Set type refers to the nature of the set that is an input and/or
output tothe fuzzy logic algorithm
Food Bioprocess Technol
Author's personal copy
-
signal. Thus, the task becomes the estimation of thisnew state
bX i (a standard task for an observer).X AX By CX Where
A 000
0mm=Y
100
24 35; B mm
xmaxxmax=Y
24 35and C 1 0 0 :An asymptotic-type observer of the system is
given as
follows:
cX 0 bX 2 k1 y0bX 0 9bX 1 m xmaxbX 1 k2 y0bX 0 10bX 2 mY XmaxbX
1 k3 y0bX 0 11where the gain vector of the observer is given
by:
K S1 CT
Si; j Si; ji j1
Each entry of the matrix S is given by the aboveequation, where
S is an n n matrix (i and j run from1 to n) and St,j are entries of
a symmetric positive
definite matrix that do not depend on . Thus, St,j aresuch that
S is a positive solution of the algebraicRiccati equation,
S A 2 I
A 2I
S CTC:
In particular, for our case, the resulting vector K is
givenby:
K k1k2k3
24 35 32mm 2 2m Ym2m3 m
" #
It is worth mentioning that we can think of this observeras a
slave system that follows the master system, whichis precisely the
real experimental system. In addition, S, asfunctional components
of the gain vector, guarantee theaccurate estimation of the
observer through the convergenceto zero of the error dynamics,
i.e., the dynamics of thedifference between the measured state and
its correspondingestimated state. One can see that generates an
extra degreeof freedom that can be tuned by the user such that
theperformance of the software sensor becomes satisfactoryfor him
(here, we use = 0.01).
The reason for the filtering effect is that the dynamicextension
acts at the level of the observer as an inte-gration of the output
of the original system (see the firstequation of the system given
by Eqs. 58 and the errorpart in the equations of system given by
Eqs. 911).The integration has averaging effects upon the
noisymeasured states. More exactly, the difference betweenthe
integral of the output of the slave part of systemand the integral
of the output of the original system
-3 -2 -1 0 1 2 3-2
-1
0
1
2
First Principal Component (74.2%)Sec
ond
Prin
cipa
l Com
pone
nt (
21.4
%)
G-30-100L
G-50-100L G-50-300L
G-40-100L
G-30-500L
G-40-300L
G-30-300L
G-50-500L
G-20-300L
Y-50-500L
Y-40-500L
Y-40-100L
B-50-100L
Y-50-100L
B-50-300LB-50-500L
B-40-300L
B-40-500LB-30-100L
Y-40-300L
Y-30-100L
B-40-100L
Y-50-300LY-30-300L
Y-30-500L
B-20-500LY-20-500L
Y-20-500LB-20-300L
B-20-100LY-20-300L
B-30-500LY-20-100L
B-30-300L
Fig. 9 Scatter plot for the first two principal components with
coloredclusters corresponding to the matrix of juice color (L, a,
and b) atdifferent process conditions: temperatures, enzymatic
complex concen-tration, and ripening stages (the three generated
clusters are marked asred triangles, blue five-pointed stars, and
green diamonds). The letters
B, Y, and G stand for brown, yellow, and green, corresponding
toripeness stages 1, 2, and 3, respectively. The percentage of the
totalvariance explained by each principal component is indicated
inbrackets
Food Bioprocess Technol
Author's personal copy
-
gives the error, and the observer is planned in such away that
the error dynamics go asymptotically to zero,which results in the
recovering of both the filtered stateand the unmeasured states.
From Fig. 10, it is possible to appreciate that thealgorithm
developed here is able to infer very accuratelythe instant juice
yield only based on the online mea-sures of torque and temperature.
Since the algorithmdoes not need to be fed with the ripeness stage,
it canbe also used to confirm the adequate selection of ba-nana
ripeness. Additionally, from Fig. 10, it is possibleto see that the
torque measurements by themselves arenot enough to detect with
precision the end of theprocess, while the prediction of juice
yield could bethe criterion to stop the enzymatic process and
continuethe next downstream step. Thus, the software
sensordeveloped here effectively rebuilds the unmeasured on-line
juice yield and also is able to filter the noisy torquesignal under
different experimental situations.
Conclusions
The optimization of the banana juice extraction processshows
that the addition of water did not significantly affectthe juice
yield, while the enzymatic complex concentrationexhibits a positive
impact on the juice yield. As such, theoptimal conditions for juice
extraction (i.e., more sustain-able) were 150 L/kg enzyme complex
concentration with-out water addition. Moreover, from this study,
it can beconcluded that the model here developed could be a
valu-able instrument for monitoring and controlling enzymaticbanana
juice extraction. Furthermore, enzyme reactors canoperate
fed-batch-wise or continuously, and the model is ahelpful tool for
optimization. Moreover, the software sensordeveloped here
effectively rebuilds the unmeasured on-linejuice yield and also is
able to filter the noisy torque signal.In addition, the
computational scheme provides a very ap-propriate tool for fast and
reliable quality control and can beused to ensure the homogeneity
of the final product. The
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.015
0.02
0.025
Time(s)
Tor
que
(Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 450020
40
60
80
Time(s)
Juic
e Y
ield
(%
)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.015
0.02
0.025
Time(s)
Torq
ue (
Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 450020
40
60
80
Time(s)
Juic
e Y
ield
(%
)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.015
0.02
Time (s)
Tor
que
(Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 4500
40
60
80
Time (s)
Juic
e Y
ield
(%
)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.015
0.02
Time (s)
Tor
que
(Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 450040
60
80
Time (s)
Juic
e Y
ield
(%
)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.0120.0140.016
Time (s)
Tor
que
(Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 4500
40
60
80
Time (s)
Juic
e Y
ield
(%
)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.01
0.012
0.014
0.016
Time (s)
Tor
que
(Nm
)
0 500 1000 1500 2000 2500 3000 3500 4000 450040
60
80
Time (s)
Juic
e Y
ield
(%
)
a) b)
c) d)
e) f)
Fig. 10 The plots show the application of the software sensor
todifferent experimental situations: in all cases, red solid lines
representthe filtered states and juice yield prediction, while blue
lines representthe noisy measured torque signal, and blue points
represent offlinejuice yield quantification. In all cases, the
enzymatic complex was
added to the 1,320 s (22 min). Ripeness stage 1 at 20 C (a) and
at30 C (b) both with 100 L of enzyme complex. Ripeness stage 2 at30
C (c) with 500 L of enzyme complex and at 50 C (d) with300 L of
enzyme complex. Ripeness stage 3 at 30 C (e) with 300 lof enzyme
complex and at 20 C (f) with 300 l of enzyme complex
Food Bioprocess Technol
Author's personal copy
-
methodology presented in this paper is general and can beused
also in automatic control applications. This supervisionscheme
provides instantaneous access to batch records, andprovides audit
tracking and traceability. This tool presentsvaluable
characteristics like real-time usage, ease of use, noneed for
sample preparation, and no sample destruction. Inaddition, PCA
proves to be a valuable tool to relate the bestvisual attributes of
the banana juice with the maturity fruitripeness stage selection
and the operation conditions, show-ing that the juice processing
temperature presents positiveeffects on juice color attributes.
Acknowledgments This work was supported by the CONACyT,Mxico
(No. 169048) and to PROMEP-SEP by funding the project.The authors
would like to thank the anonymous reviewers for theirvaluable
comments and suggestions to improve the document.
References
Aguilar, C. N., Gutirrez-Snchez, G., Rado-Barragn, P. A.,
Rodrguez-Herrera, R., Martnez-Hernandez, J. L., &
Contreras-Esquivel, J. C.(2008). Perspectives of solid state
fermentation for production offood enzymes. American Journal of
Biochemistry and Biotechnol-ogy, 4(4), 354366.
Akubor, P. I. (1996). The suitability of African bush mango
juice forwine production. Plant Foods for Human Nutrition, 49,
213219.
Bahramian, S., Azin, M., Chamani, M., & Gerami, A. (2011).
Optimi-zation of enzymatic extraction of sugars from Kabkab date
fruit.Middle East Journal of Scientific Research, 7(2), 211216.
Boudhrioua, N., Giampaoli, P., & Bonazzi, C. (2003). Changes
inaromatic components of banana during ripening and
air-drying.LWTFood Science and Technology, 36(6), 633642.
Bradford, M. M. (1976). A rapid and sensitive method for the
quanti-tation of microgram quantities of protein utilizing the
principle ofproteindye binding. Analytical Biochemistry, 72,
248254.
Buenrostro-Figueroa, J., Garza-Toledo, H., Ibarra-Junquera, V.,
&Aguilar, C. N. (2010). Juice extraction from mango pulp
usingan enzymatic complex of Trichoderma sp produced by
solid-statefermentation. Journal of Food Science and Biotechnology,
19(5),13871390.
Casimir, D. J., & Jayaraman, K. S. (1971). Banana drink, a
new cannedproduct. CSIRO Food Research Quarterly, 31(1/2),
2429.
Chvez-Rodrguez, A. M., Prez-Martnez, J. D., Ibarra-Junquera,
V.,Escalante-Minakata, P., VillaVelzquez, C. I.,
Dibildox-Alvarado,E., & Ornelas-Paz, J. de J. (2013).
Microencapsulation of bananajuice from three different cultivars.
International Journal of FoodEngineering. 9(1), 18.
Cheirsilp, B., & Umsakul, K. (2008). Processing of
banana-based wineproduct using pectinase and -amylase. Journal of
Food ProcessEngineering, 31(1), 7890.
Combo, A. M. M., Aguedo, M., Goffin, D., Wathelet, B., &
Paquot, M.(2012). Enzymatic production of pectic oligosaccharides
frompolygalacturonic acid with commercial pectinase
preparation.Food and Bioproducts Processing, 90(3), 588596.
Cozzolino, D., Cynkar, W., Shah, N., & Smith, P. (2011).
Technicalsolutions for analysis of grape juice, must, and wine: the
role ofinfrared spectroscopy and chemometrics. Analytical
andBioanalytical Chemistry, 401, 14751484.
ENMEX, S.A. de C.V. (2003) Macerex PM: Pectinase and
CellulaseEnzymatic Complex to be applied in fruit maceration.
Tlalnepantla,
Edo. De Mxico. Available at: www.enzymes.com.mx/awsHome.htm.
Accessed 21 October 2011
Escalante-Minakata, P., Ibarra-Junquera, V., Gonzlez-Garca,
R.,Rosu, H. C., & de Len-Rodrguez, A. (2009). On-line
estimationof kinetic rates in Mezcal fermentation based on redox
measure-ments. Bioprocess and Biosystems Engineering, 32, 4752.
Escalante-Minakata P., Ibarra-Junquera V., Chvez-Rodrguez
A.M.,Ornelas-Paz, J.J., Emparan-Legaspi M.J., Prez-Martnez
J.D.,Villavelzquez C., (2013). Evaluation of the freezing and
thawingcryoconcentration process on bioactive compounds present
inbanana juice from three different cultivars. To appear in
Interna-tional Journal of Food Engineering (in press).
Fernndez-Fernndez, M., & Prez-Correa, R. (2010). Solid
substratefermentation, automation. In M.C. Flickinger (Ed.),
Encyclopediaof industrial biotechnology, bioprocess, bioseparation,
and cell tech-nology. New York: Wiley. ISBN: 978-0-471-79930-6. 7
Volume Set
Falade, K. O., & Babalola, S. O. S. (2004). Degradation of
qualityattributes of sweetened Julie and Ogbomoso mango juices
duringstorage. European Food Research and Technology, 218,
456459.
Hirst, K., Zerbe, G. O., & Hay, W. W., Jr. (1996). A
generalizedMichaelisMenten response surface. Statistics in
Medicine, 15,21072119.
Ibarra-Junquera, V., Escalante-Minakata, P., Mancilla-Margalli,
N. A.,Murgua, J. S., Garca dela Rosa, L. A., & Rosu, H. C.
(2010).Strategies to monitor alcoholic fermentation processes. In
K. Jurgen& F. Oswald (Eds.), Industrial fermentation: food
processes, nutrientsources and production strategies. Hauppauge,
NY: Nova Science.
Ibarra-Junquera, V., Torres, L. A., Rosu, H. C., Argello, G.,
&Collado-Vides, J. (2005). Nonlinear software sensor
monitoringgenetic regulation processes with noise and modeling
errors.Physics Review E, 72(1), 011919.
Khalil, K. E., Ramakrishna, P., Nanjundaswamy, A. M.,
&Patwardhan, M. V. (1989). Rheological behavior of
clarifiedbanana juice: effect of temperature and concentration.
Journal ofFood Engineering, 10, 231240.
Krener, A. J., & Isidori, A. (1983). Linearization by output
injection andnonlinear observers. Systems and Control Letters,
3(1), 4752.
Kiyoshi, M., & Wahachiro, T. (2003). Change of polyphenols
com-pounds in banana pulp during ripening. Food Preservation
Sci-ence, 29(6), 347351.
Kyamuhangire, W., & Pehrson, R. (1999). Conditions in banana
rip-ening using the rack and pit traditional methods and their
effect onjuice extraction. Journal of the Science of Food and
Agriculture,79, 347352.
Kyamuhangire, W., Myhre, H., Sorensen, H. T., & Pehrson, P.
(2002).Yield, characteristics and composition of banana juice
extractedby the enzymatic and mechanical methods. Journal of the
Scienceof Food and Agriculture, 82, 478482.
Lee, W. C., Yusof, S., Hamid, N. S. A., & Baharin, B. S.
(2006a).Optimizing conditions for hot water extraction of banana
juiceusing response surface methodology (RSM). Journal of
FoodEngineering, 75, 473479.
Lee, W. C., Yusof, S., Hamid, N. S. A., & Baharin, B. S.
(2006b).Optimizing conditions for enzymatic clarification of banana
juiceusing response surface methodology (RSM). Journal of
FoodEngineering, 73, 5563.
Lee,W. C., Yusof, S., Hamid, N. S. A., &Baharin, B. S.
(2007). Effects offining treatment and storage temperature on the
quality of clarifiedbanana juice. LWTFood Science and Technology,
40, 17551764.
Lpez-Nicols, J. M., Prez-Lpez, A. J., Carbonell-Barranchina,
A.,& Garca-Carmona, F. (2007). Kinetic study of the activation
ofbanana juice enzymatic browning by the addition of
maltosyl--cyclodextrin. Journal of Agricultural and Food Chemistry,
55,96559662.
Luenberger, D. (1971). An introduction to observers. IEEE
Transac-tions on Automatic Control, 16(6), 596602.
Food Bioprocess Technol
Author's personal copy
http://www.enzymes.com.mx/awsHome.htmhttp://www.enzymes.com.mx/awsHome.htm
-
Mohapatra, D., Mishra, S., Singh, C. B., & Jayas, D. S.
(2011). Post-harvest processing of banana: opportunities and
challenges. Foodand Bioprocess Technology, 4(3), 327339.
Onwuka, U. N., & Awam, F. N. (2001). The potential for
bakers yeast(Saccharomyces cerevisiae) in the production of wine
from ba-nana, cooking banana and plantain. Food Service Technology,
1,127132.
Reddy, L. V. A., & Reddy, O. V. S. (2005). Production and
character-ization of wine from mango fruit (Mangifera indica L).
WorldJournal of Microbiology and Biotechnology, 21, 13451350.
Sinija, V. R., & Mishra, H. N. (2011). Fuzzy analysis of
sensory datafor quality evaluation and ranking of instant green tea
powder andgranules. Food Bioprocess Technology, 4, 408416.
Sreenath, H. K., Sudarshanakrishna, K. R., & Santhanam, K.
(1994).Improvement of juice recovery from pineapple pulp/residue
usingcellulases and pectinases. Journal of Fermentation and
Bioengi-neering, 78(6), 486488.
Srivatsa, G.S. (1996). Biotechnology products. In C.M. Riley,
T.W.Rosanske (Eds.), Development and validation of
analyticalmethods, volume 3 (first edition). Oxford: Elsevier.
Viquez, F., Lastreto, C., & Cooke, R. D. (1981). A study of
theproduction of clarified banana juice using pectinolytic
enzymes.Journal of Food Technology, 16, 115122.
Virgen-Ortz, J. J., Ibarra-Junquera, V., Osuna-Castro, J. A.,
Escalante-Minakata, P., Mancilla-Margalli, N. A., & de
Ornelas-Paz, J.(2012). Method to concentrate protein solutions
based on dialy-sisfreezingcentrifugation. Analytical Biochemistry,
426(1), 412.
Wall, M. M. (2006). Ascorbic acid, vitamin A, and mineral
composi-tion of banana (Musa sp.) and papaya (Carica papaya)
cultivarsgrown in Hawaii. Journal of Food Composition and Analysis,
19,434445.
Zawojewski, J. (2010). Problem solving versus modeling. In R.
Lesh,P. Galbraith, C. R. Haines, & A. Hurford (Eds.), Modeling
stu-dents mathematical modeling competencies: ICTMA 13 (pp.237244).
New York: Springer.
Zhang, H. (2009). Software sensors and their applications
inbioprocess. In M.C. Nicoletti & L.C. Jain (Eds.),
Computationalintelligence techniques for bioprocess modelling (pp.
2556).Berlin: Springer
Zhong, J. J. (2010). Recent advances in bioreactor engineering.
KoreanJournal of Chemical Engineering, 27(4), 10351041.
Zolfaghari, M., Sahari, M. A., Barzegar, M., & Samadloiy, H.
(2010).Physicochemical and enzymatic properties of five kiwifruit
culti-vars during cold storage. Food and Bioprocess Technology,
3,239246.
Food Bioprocess Technol
Author's personal copy
Optimization, Modeling, and Online Monitoring of the Enzymatic
Extraction of Banana
JuiceAbstractIntroductionMethodologyBioreactorBiological
MaterialJuice Extraction ProcedureJuice YieldProtein
QuantificationColorimetric AnalysisStatistical AnalysisVisual
EvaluationOptimization ProcedureFuzzy Logic ApproachEnzymatic Juice
Extraction ModelSoftware Sensor Design
Results and DiscussionOptimization ProcedureModeling the
Enzymatic Juice ExtractionRemarks on Enzyme Activity and
TemperatureColorimetric AnalysisThe Software Sensor
ConclusionsReferences