Journal of Food Engineering 192 (2017) 61e71

Contents lists available at ScienceDirect

Journal of Food Engineering

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

Dynamic modelling of pectin extraction describing yield and

functional characteristics
N.M. Andersen a, T. Cognet b, P.A. Santacoloma c, J. Larsen c, I. Armagan c, F.H. Larsen d,
K.V. Gernaey a, J. Abildskov a, J.K. Huusom a, *
a
CAPEC-PROCESS Research Center, Department of Chemical and Biochemical Engineering, Technical University of Denmark, SøltoftsPlads, Building 229,
2800 Kgs. Lyngby, Denmark
b
ENSIACET, National Polytechnic Institute of Chemical and Industrial Engineering and Technology, 4, all
ee Emile Monso, 31030 Toulouse Cedex 4, France
c
CPKelco, Ved Banen 16, 4623 Lille Skensved, Denmark
d
Department of Food Science, University of Copenhagen, 1958 Frederiksberg, Denmark

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

Article history: A dynamic model of pectin extraction is proposed that describes pectin yield, degree of esterification and
Received 26 May 2016 intrinsic viscosity. The dynamic model is one dimensional in the peel geometry and includes mass
Received in revised form transport of pectin by diffusion and reaction kinetics of hydrolysis, degradation and de-esterification. The
9 August 2016
model takes into account the effects of the process conditions such as temperature and acid concen-
Accepted 11 August 2016
tration on extraction kinetics. It is shown that the model describes pectin bulk solution concentration,
Available online 12 August 2016
degree of esterification and intrinsic viscosity in pilot scale extractions from lime peel at different
temperatures (60
C, 70
C, 80
C) and pH (1.5, 2.3, 3.1) values.
Keywords:
Pectin extraction
© 2016 Elsevier Ltd. All rights reserved.
Dynamic modelling
Yield
Degree of esterification
Intrinsic viscosity

1. Introduction material as it is the residual product from juice production.

Pectin is a family of complex polysaccharides primarily present
in plant cell walls. The three main components of pectin are
homogalacturonan (HG), rhamnogalacturonan I (RG-I) and rham-
nogalacturonan II (RG-II). From an industrial point of view HG is the
most interesting part of pectin and is composed of a chain of a-1,4
linked galacturonic acid residues. Some of the residues may be
acetylated or esterified/methoxylated (Willats et al., 2006). Pectin
is widely used in many food and pharmaceutical products for
modification of rheological properties. The worldwide market for
pectin exceeds 45,000 tonnes and represents a value of more than
400 million EUR (Savary et al., 2003). Citrus peel from lemon, or-
ange and lime is the main source of raw material for industrial
production of pectin and citrus peel is also a convenient raw
* Corresponding author.
E-mail address: jkh@kt.dtu.dk (J.K. Huusom).
Industrially, pectin is extracted i.e. leached from citrus peels by acid
hydrolysis, and subsequently recovered by precipitation with
alcohol (Graham and Sheperd, 1953; Minkov et al., 1996).
A pectin factory receives shipments of peels of different citrus
fruits from various geographical locations and therefore there is a
variation in properties and chemical composition of the raw ma-
terial. This variation requires a constant adjustment of the process
conditions in order to obtain not only a high pectin yield but also
the desired product characteristics. Some of the key pectin char-
acteristics are degree of esterification (DE) and intrinsic viscosity
(IV) (Pagan et al., 1999). The target values of these depend on the
specific product application.
During extraction pectin yield increases, when temperature in-
creases and pH decreases but temperature and pH also affect DE
and IV. Therefore, these process parameters as well as the extrac-
tion time must be optimized in order to achieve an optimal
compromise between pectin yield and characteristics (Sirisakulwat

http://dx.doi.org/10.1016/j.jfoodeng.2016.08.006
0260-8774/© 2016 Elsevier Ltd. All rights reserved.
62 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
et al., 2008).
Statistical models of optimized pectin extraction with respect to
yield and DE by response surface methodologies have been re-
ported by numerous researchers, for instance Wai et al. (2009),
Masmoudi et al. (2008) and Voragen et al. (2003). First principles
models (also referred to as physics-based or white-box models) to
predict pectin yield have been studied by Minkov et al. (1996),
Panchev et al. (1989) and Dur an et al. (2015) in lab scale at con-
stant temperature and pH. Cho and Hwang (2000) combined a ki-
netic study of pectin extraction with a model of intrinsic viscosity
including the influence of temperature on extraction rate constants.
The influence of temperature and pH on extraction kinetics has
been studied by Pag an and Ibarz (1999). They correlated yields with
pH using a third order polynomial. The kinetics of the de-
esterification of pectin has been studied by Kirtchev et al. (1989)
for extractions at constant pH. There is in the literature a lack of a
general first principles model for pectin extraction which couples
the reaction and transport phenomena determining the yield of
extraction with the functional characteristics of the extracted
pectin through IV and DE. For such a model to create value in
relation to commercial pectin production methods such a model
should describe both the microscopic phenomena at the peel level
as well as the overall behavior that can be observed in an extraction
tank at scale.
This paper presents a dynamic one-dimensional model of a
batch pectin extraction process in order to predict pectin concen-
tration, DE and IV when temperatures and pH vary. The model
includes diffusion, reaction kinetics and mass transfer. The rate
constants are functions of temperature and pH. A methodology for
efficient estimation of the process parameters is proposed and the
model is fine-tuned against pilot scale data generated based on a
central composite experimental design. In this sense the proposed
contribution is in line with the increased focus on first principles
modelling and virtualizations in food engineering from the Vir-
ProFood meeting in Salerno (2014) and a recent special issue (Vol.
176, 2016) in Journal of Food Engineering as pointed out in the
editorial by Marra (2016) and the opening paper by Saguy (2016).
The paper is organized with model development and the parameter
estimation strategy in the following section, materials and methods
in section 3 and results and discussion in section 4 and 5 followed
by concluding remarks in section 6.
2. Modelling
In this section a dynamic model of the pectin extraction process
is described. The purpose is to describe the main phenomena
involved in a batch extraction reactor to predict the effect on yield
and pectin properties of changes in the key process conditions:
temperature and pH. Balance equations using first principles are
utilized to the extent possible to describe phenomena that can be
observed at the reactor level. The following simplifying assump-
tions are made:
1. The peel is considered as a structure where the active compo-
nent (water insoluble protopectin) is embedded.
2. Water penetrates immediately into the inside of the peel and
this leads to release of pectin.
3. Initial concentrations of protopectin and pectin are uniformly
distributed in the raw peel.
4. Global reaction and transport coefficients are assumed for the
peel.
5. The peels have the shape of flakes with equal size, and are
described using a one-dimensional slab geometry.
6. The peel slab geometry is unaffected by the penetration of
water.
7. Pectin in the bulk solution is homogenously dispersed as this
phase is considered to be well mixed.
The raw peels have been washed with water prior to shipping
and during this process pectin has been released from the peels.
In the subsequent drying process it is assumed that some of this
pectin is re-captured in the peel. Therefore addition of water to
the peels results in an initial concentration of water soluble
pectin prior to extraction. It is known that the peel morphology
consist of several layers which could influence the reaction and
mass transport. It is however assumed that a lumped approach to
model these as global coefficients for the peel is sufficient in the
development of a model for the overall behavior of the leaching
process.
The peel flakes are modelled by a one-dimensional slab geom-
etry with a thickness of 2L. The origin of the x-axis (i.e. x ¼ 0) is
placed in the centre of the flake so that the exposed surfaces of the
flake are respectively at x ¼ -L and x ¼ L (see Fig. 1). The flake is
assumed to be symmetric around the centre, and the boundary
conditions at x ¼ -L and x ¼ L are identical, therefore the concen-
tration profiles are also symmetric around x ¼ 0. This means that
c(x) ¼ c (-x) and therefore the system only needs to be solved in the
interval from x ¼ 0 to x ¼ L.
In the peel, acid catalyzed hydrolysis of protopectin takes place
and there is internal diffusion of pectin. The proton concentration
in the solution is adjusted through addition of a strong acid. There is
a mass transport of pectin through the boundary layer to the bulk
solution. In the bulk solution, pectin is degraded and ester groups
on pectin in the bulk solution may be converted into free acid form.
A control volume of thickness Dx along the full height and width of
the flake is placed inside the peel between x ¼ 0 and x ¼ L in order
to describe spatial variation in concentrations in the peel. The acid
catalyzed hydrolysis of protopectin:
½H þ 
ProtopectinðpeelÞ



!PectinðpeelÞ
is assumed to be a first order reaction and therefore this leads to the
differential equation:
vcprotopectinðpeelÞ ðt; xÞ  h i
¼ khydrolysis T; Hþ $cprotopectinðpeelÞ ðt; xÞ
vt
(1)
wherecprotopectin(peel) is the concentration of protopectin in the peel
and khydrolysis(T,[Hþ]) is the reaction rate of hydrolysis taking into
account the temperature and pH dependency. The diffusion of
Fig. 1. Batch reactor with peels and the 1D flake model of a single peel.
 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
pectin in the peel is described by Fick's first law (Seader and Henley,
2006). The pectin balance inside the peel leads to the following
partial differential equation for the concentration of pectin in the
peel:
vcpectinðpeelÞ ðt; xÞ v2 cpectinðpeelÞ ðt; xÞ
¼ Dpectin ðTÞ
vt vx2
 h i
þ khydrolysis T; Hþ $cprotopectinðpeelÞ ðt; xÞ
where cpectin(peel) is the concentration of pectin in the peel and
Dpectin(T) is the diffusion coefficient as function of temperature of
pectin in the wet peel material. Equation (2a) requires two
boundary conditions. The diffusion flux at x ¼ 0 must be zero
because of symmetry. Then this leads to the following Neumann
boundary condition:
vcpectinðpeelÞ
¼0 at x ¼ 0
vx
The transport of pectin from the peel to the solution is given by
the rate of mass transfer of pectin through the boundary layer. In
this case mpectin follows the general mass transfer model which is
expressed in terms of a mass transfer coefficient,kmasstransfer the area
available for the mass transfer, ATotal and the driving force given by
the difference between the concentration of pectin at the peel
surface, cpectin(peel)(t,L) and in the bulk solution, cpectin(bulk)(t)(Min-
kov et al., 1996):
 
mpectin ¼ ATotal $kmasstransfer $ cpectinðpeelÞ ðt; LÞ  cpectinðbulkÞ ðtÞ
At the surface of the peel the rate of mass transfer of pectin by
diffusion to the peel surface must be equal to the rate of mass
transfer through the boundary layer and into the solution (Seader
and Henley, 2006). This leads to the following boundary condition:
vcpectinðpeelÞ kmasstransfer 
¼  cpectinðpeelÞ ðt; LÞ
vx Dpectin
 esterification in the bulk solution.
 cpectinðbulkÞ ðtÞ at x
¼L
In the bulk solution, pectin may undergo degradation re-
actions due to the temperature and acidity in the media.
Consequently, this is directly reflected in a lower molecular
weight of the pectin molecule, i.e. the molecular weight distri-
bution of pectin will be shifted towards lower chain lengths. Cho
and Hwang (2000) suggest first order kinetic for this process
assuming an average chain length for the pectin just being
extracted and another representative for the degraded pectin. A
pectin balance including transfer through the boundary layer and
degradation leads to the following differential equation for
pectin bulk solution concentration:
vcpectinðbulkÞ ðtÞ ATotal $kmasstransfer 
¼ $ cpectinðpeelÞ ðt; LÞ
vt VTotal
 change in intrinsic viscosity (IV), k0,i is the Arrhenius pre-
 cpectinðbulkÞ ðtÞ
 h i
 kdegradation T; Hþ $cpectinðbulkÞ ðtÞ
whereATotal is the total surface area of flakes, VTotal is the total vol-
ume of liquid in the batch reactor and kdegradation(T,[Hþ]) is the re-
action rate constant for the degradation reaction as function of
temperature and pH. The concentration of degraded pectin in the
63
bulk solution is described by the differential equation:
vcdegradedpectinðbulkÞ ðtÞ  h i
¼ kdegradation T; Hþ $cpectinðbulkÞ ðtÞ (4)
vt
A certain fraction, fGA of pectin consists of galacturonic acid
residues. The degree of esterification is the ratio of esterified gal-
acturonic acid residues to the total amount of galacturonic acid
residues (Sriamonsak, 2003). The total amount of galacturonic acid
includes galacturonic acid of degraded pectin because degraded
(2a) pectin contains shorter chains of linked galacturonic acid but the
degradation reaction is not assumed to change the fraction of gal-
acturonic acid in the pectin structure. The degree of esterification,
DE is calculated as:
cesterðbulkÞ ðtÞ
DEðtÞ ¼  (5)
fGA $cpectinðbulkÞ ðtÞ þ cdegradedpectinðbulkÞ ðtÞ
During extraction, some of the ester groups on pectin in the bulk
(2b) solution are converted into free acid form. According to Kirtchev
et al. (1989) this de-esterification reaction can be described as a
first-order reaction. A balance of the degree of esterification of the
extracted pectin in the bulk solution leads to the following differ-
ential equation:
vcesterðbulkÞ ðtÞ ATotal $kmasstransfer 
¼ DE0 $fGA $ $ cpectinðpeelÞ ðt; LÞ
vt VTotal

 cpectinðbulkÞ ðtÞ
 h i
 kdeesterification T; Hþ $cesterðbulkÞ ðtÞ
(6)
where DE0 is the initial degree of esterification, i.e. inherent to the
raw material, and kdeesterification ðT; ½H þ Þ is the reaction rate con-
stant as function of temperature and pH for the de-esterification
process. The first term on the right hand side of equation (6) de-
scribes the contribution of ester groups of pectin transferred from
the peel to the bulk solution. The second term describes de-
The model presented in Equations (1)e(6) contains diffusion
and reaction terms. The diffusion terms depend on the tempera-
(2c)
ture and the reaction terms depend on process conditions such as
temperature and pH. These dependencies can be described by
Arrhenius type equations which for the diffusion coefficient are
expressed as in Seader and Henley (2006):
Dpectin ¼ D0 expð  Da =RTÞ (7)
where D0 is the pre-exponential factor and Da is a coefficient which
can be interpreted as an energy barrier for the diffusion process. R is
the gas constant and T is temperature. For the dependencies in the
reaction rate constants, the Arrhenius equation is used:
  
ki ¼ k0;i exp  Ea;i ðRTÞ (8a)
where i denotes hydrolysis, degradation, de-esterification or
exponential factor andEa,i is the energy of activation for the i'th
reaction. The pre-exponential factor k0,i depends on the type of
(3) catalyst and its concentration, which in this case related to the pH
in the batch reactor as the acid acts as a homogenous catalyst for
the reaction. The concentration of protons and acid is approxi-
mately constant during the batch reaction when a strong acid such
as nitric acid is used. This pre-exponential factor is therefore
expressed as a linear function of the proton concentration
64 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
h i
k0;i ¼ ai $ Hþ þ bi
The coefficients ai and bi will need to be fitted to experimental
data. This functional expression was also used by Girisuta et al.
(2007) for acid-catalyzed hydrolysis of cellulose.
The initial conditions for the model described by Equations
(1)e(6) are:
cprotopectin ð0; xÞ ¼ c0protopectin
cpectinðpeelÞ ð0; xÞ ¼ c0pectinðpeelÞ
cpectinðbulkÞ ð0; xÞ ¼ 0
cdegradedpectinðbulkÞ ð0; xÞ ¼ 0
cestergroupsðbulkÞ ð0; xÞ ¼ 0
The extracted soluble pectin can undergo a degradation reaction
in the warm aqueous acid solution which results in pectin of shorter
chain length and hence reduced gelling properties. For industrial
production this is monitored using analysis of the intrinsic viscosity
of the extract. The intrinsic viscosity, IV is defined as (Cho and
Hwang, 2000):
1 ðh  hs Þ
IV ¼ lim
c/0 c hs
where hs is the viscosity of the solvent, h is the viscosity of the
solution and c is the concentration of the solute. The intrinsic vis-
cosity of pectin is affected by for instance molecular weight and pH
(Sriamonsak, 2003). The change in intrinsic viscosity during
extraction is modelled as a first order reaction (Cho and Hwang,
2000), which is independent of the balance equations described
in equations (1)e(6), as this is a good description of what is
observed in the extraction process.
The intrinsic viscosity is modelled as:
IV ¼ IV0 $expð  kIV $tÞ
where IV0 is the initial intrinsic viscosity, t is time and kIV is the
reaction constant. The influence of the process temperature and pH
on the extraction process at reactor level is described as in the
Arrhenius function (8a) and (8b).
The proposed model for the IV is not an explicit part of the first
principle model which describes the yield and DE of the extracted
pectin in the bulk solution. It would have been necessary to model
the polymer size distribution in the bulk solution using a popula-
tion balance model in order to infer IV and it is evaluated that this
level of detail cannot be justified in relation to the process analytics
which is implemented on our pilot plant or in commercial pectin
production.
2.1. Parameter estimation strategy
In order to estimate the parameters in the model described in
equations (1)e(6) and the model described in equation (9), it is
necessary to have transient data of pectin bulk solution concen-
tration, DE and IV for the extractions and to make perturbations of
temperature and acid concentration.
The parameters that have to be estimated in the model described
in equations (1)e(6) are: fD0 ; Da ; kmasstransfer ; ATotal ; DE0 ; Ea;hydrolysis
; ahydrolysis ; bhydrolysis ; Ea;degradation ; adegradation ; bdegradation ;
Ea;deesterification adeesterification ; bdeesterification g and the
(8b) initial concentrations c0protopectin c0pectinðpeelÞ .
The parameters that have to be estimated in the model
described in equation (10) are: {Ea;IV ,aIV ,bIV ,IV0 }.
The total surface area of flakes, ATotal, can be roughly calculated
based on the total volume of peel Vpeel, the volume of one flake
Vflake, and the area of one side of a single flake Aflake by the equation:
Vpeel
ATotal ¼ $A $2
Vflake flake
The first fraction estimates the number of flakes whereas the
twice the area of a flake is an estimate for its total surface area. The
rest of the parameters can be divided into four groups:
1. Parameters concerning diffusion and mass transfer of pectin.
2. Parameters concerning the hydrolysis and degradation reactions
of pectin.
3. Parameters concerning the de-esterification reaction of pectin.
4. Parameters concerning the intrinsic viscosity of pectin.
Measurements of pectin bulk solution concentration are used to
estimate parameters in group 1 and 2. During extraction diffusion,
mass transfer and reactions occur simultaneously. In order to
separate these effects, two different extractions are performed:
extractions without acid (water extractions) and extractions with
acid. The extractions without acid are used for estimating the pa-
rameters in group 1 because it is assumed the hydrolysis reaction
(9)
without acid is slow enough to be neglected during a short
extraction and that the degradation reaction can also be neglected
in this timespan where the pectin concentration in the bulk solu-
tion is relative small. This means khydrolysis ¼ 0 and kdegradaation ¼ 0.
In this extraction, pectin in the bulk solution results only from
diffusion and mass transfer of the initial pectin in the peel. There-
fore the parameters {c0pectinðpeelÞ ,kmasstransfer,Da,D0} can be estimated
from measurements of pectin concentration in the extractions
without acid. We have chosen data from extractions in up till
120 min for this estimation in order to eliminate the influence of
potentially slow hydrolysis or degradation reactions which could
data from long extractions.
(10) The extractions with acid are used for estimating the parameters
in group 2. In these extractions, hydrolysis and degradation take
place and the measurements of pectin concentration are used for
estimating the parameters {c0protopectin ,Ea,hydrolysis,ahydrolysis,bhydrolysis,
Ea,degradation,adegradation,bdegradation}.
The parameters in group 3 that concern the de-esterification of
pectin {Ea,deesterification,adeesterification,bdeesterification,DE0} are esti-
mated from a series of measurements of DE. These parameters can
be estimated separately because they only appear in the equation
for the concentration of ester groups in the bulk solution. Hence
from the model it can be seen that the balance of the esterified
galacturonic acid groups in the bulk solution does not affect the
actual concentration of pectin in the bulk solution.
The parameters in group 4 that concern the intrinsic viscosity of
pectin {Ea,IV,aIV,bIV,IV0} are estimated from a series of measure-
ments of IV.
Strong parameter correlation between the pre-exponential
factor and the activation energy in Arrhenius expressions such as
in (8) are well known and undesired in relation to determining
these from experimental data. This correlation can be reduced by
centering the expression on a temperature T0 (Himmelblau, 1970).
     
ki ¼ k0;i exp  Ea;i ðRT0 Þ $exp  Ea;i R$ð1=T  1=T0 Þ
This expression is used in the parameter estimation of pre-
 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
Fig. 2. Overview of the central composite design for the pectin extractions.
exponential factors and activation energies in (7) and (8).
Initial parameter guesses are necessary in order to perform
parameter estimation. In section 3.4 qualified initial parameter
guesses for the model are described.
3. Materials and methods
3.1. Design of experiments
A central composite experimental design has been made to
study the effects of temperature and pH on the extraction kinetics
Fig. 3. PI-diagram (left) and picture (right) of the pilot plant setup for pectin extraction.
65
as shown in Fig. 2. Fourteen experiments have been performed with
four replicates at the center points (T ¼ 70
C, pH ¼ 2.3) and one
extraction at each of the other points. High pectin yield is favoured
by values close to pH ¼ 1 whereas high DE and IV are favoured by
values close to pH ¼ 4 (Sirisakulwat et al., 2008). Therefore a pH
range from 1.5 to 3.1 is selected with pH ¼ 2.3 as the center point. A
temperature range from 60
C to 80
C has been selected as ex-
tractions at high temperatures induce high pectin yields (Paga n
et al., 1999). In addition three extractions without acid were per-
formed at temperatures of 60, 70 and 80
C respectively.
3.2. Pectin extractions in pilot plant
The experimental setup consists of two tanks T1 and T2 (see
Fig. 3). In tank T1 16 L of demineralized water is heated by the
heating element HE1. Heated water is transferred to tank T2 where
it is mixed with peel and acid. Temperature is controlled using the
heating element HE2. The extraction takes place in tank T2 for
270 min under continuous stirring. Samples are taken every 30 min
from V3.
Pectin was extracted from 0.5 kg of dried lime peel with 16 L of
demineralized water and with 70, 300 or 900 mL 1M nitric acid
added and at temperatures of 60, 70 or 80
C respectively. The
samples were filtrated. The filtrate was analyzed by flow injection
polymer analysis (FIPA) in order to measure pectin concentration
and intrinsic viscosity whereas DE was determined by 1H NMR
spectroscopy and chemometrics according to the procedure in
Winning et al. (2007).
3.3. Flow injection polymer analysis
Pectin polymer content and intrinsic viscosity of pectin poly-
mers were determined by size exclusion chromatography (SEC)
66 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71

Table 1 estimates. These information can be use to approximate the

Process inputs and parameters for the pectin extractions.
Process parameter Value
Slab thickness 2L ¼ 0.002 m
Slab cross-sectional area A ¼ 0.01  0.01 m2
Total mass of lime peel m ¼ 0.5 kg
Fraction of galacturonic acid in lime peel fGA ¼ 0.84
Total volume of water V ¼ 0.016 m3
Temperature T ¼ 60
C, 70
C, 80
C
pH pH ¼ 1.5, 2.3, 3.1
coupled with a triple detection array (TDA 305, Viscotek Corp.,
Houston, USA). The TDA consisted of a refractive index (RI) detector,
two light scattering detectors 90
and 7
, in addition to an online
viscometer. The FIPA software converts the detector signals to
intrinsic viscosity and concentration.
Prior to injections, samples were centrifuged at 10000 g for
10 min and 1 ml of the supernatant was diluted 10 times with 0.3 M
lithium acetate (pH 4.6). Samples were injected using an integrated
Viscotek GPCMax autosampler (VE, 2001; Viscotek, Oss, The
Netherlands). 0.3 M lithium acetate (pH 4.6) was used as the eluent
with a flow rate of 1 mL/min at 37
C. The separation was carried
out using a SEC 60 (150  7.8 mm) column (Thermo Fisher Scien-
tific, MA, USA).
3.4. Computational method
The set of partial differential equations, PDE's, for cpectin(peel) in
the model is transformed into a set of ordinary differential equa-
tions, ODEs, by discretization of the spatial derivatives using central
differences. The resulting system of coupled differential equations
is implemented in Matlab® and solved numerically in time in a
method of lines approach using ode15s for stiff systems (Shampine
and Reichelt, 1997) with the process parameters given in Table 1. If
the model equations are to be solved simultaneously in both space
and time though a finite difference scheme we would recommend
either the implicit method or the Crank-Nicolson methods for
ensuring numerical robustness of the integration scheme. The
Matlab® function lsqnonlin for nonlinear least-squares problems is
used for parameter estimation. This function returns the parameter
estimates as well as the residuals for the regression and the
gradient of the least squares objective function with respect to the
covariance matrix for the parameter estimates and evaluation of
confidence limits. A common problem for nonlinear least-squares
problems is that the minimum found may be a local minimum, so
the solution depends on the quality of the initial guesses of the
parameters. In the following qualified initial parameter guesses for
the model are described.
The density of wet citrus peel is 1030 kg/m3(Zhou et al., 2008).
For mpeel ¼ 0.5 kg this gives the total volume of peel
Vpeel ¼ 4.85.104 m3. The initial mass of protopectin and pectin in
the peel is calculated by multiplying the total mass of peel and a
theoretical conversion factor, c:
mprotopectinþpectin ¼ mpeel $c
The pectin content of citrus peel is usually 20e30% of the dried
peel mass (Kanmani et al., 2014), therefore a conversion factor of
0.25 is selected as initial guess. This gives mprotopectinþpectin¼0.125
kg. As the distribution of protopectin and hydrolysed pectin is
unknown, it is assumed that half of the 0.125 kg is hydrolysed. This
gives the following crude initial guess of the concentrations:
c0pectinðpeelÞ ¼ c0protopectinðpeelÞ ¼ 129kg=m3
In the model of Cho and Hwang (2000) the diffusion and hy-
drolysis reaction were considered as one phenomenon and
described by a single hydrolysis rate constant with a corresponding
activation energy of 17.8$103 J/mol. The order of magnitude for
khydrolysis is 104s1 (Minkov et al., 1996), which means that order of
magnitude of the pre-exponential factor, k0,hydrolysis, and initial
guesses for ahydrolysis and bhydrolysis used to model this pre-
exponential factor can be estimated and use as initial guess in the
parameters estimation. The same analysis can be used for the pa-
rameters related to the degradation reaction where Cho and Hwang
(2000) reports that the energy of activation for the reaction rate
constant for degradation is 62.7$103 J/mol and the order of
magnitude for kdegradation is 105s1.
The initial guesses of parameters for the model of intrinsic vis-
cosity given in equation (9) are found by plotting ln (IV) as function
of time for each temperature and pH value. This gives IV0 around
0.76 L/g and kIV is of the order 105 s1. The initial guess of Ea,IV is
found by plotting ln (kIV) as function of 1/T. This gives Ea,IV is of the
order 104 J/mol and k0,IV is of the order 103 s1. Therefore the initial
guesses for aIV and bIV in the expression for the pre-exponential

Table 2
Values of model parameters found by parameter estimation, 95% confidence intervals, and initial values for the parameter estimation found in the literature.

Parameter Estimation 95% confidence intervals Initial value from literature

c0pectinðpeelÞ 77.3 kg/m3 [75.1;79.4]

c0protopectinðpeelÞ 180 kg/m3 [178;181]

kmasstransfer 1.10$103 m/s [0.10$103;2.10$103] 6.62$105 m/s (Pagan et al., 1999)
Da 49.1$103 J/mol [41.0$103;57.2$103] 19.1$103 J/mol (Genchev et al., 1988)
D0 1.70$103 m2/s [1.36$103;2.04$103] 3 2
1.13$10 m /s (Genchev et al., 1988)
Ea,hydrolysis 81.1$103 J/mol [71.4$103;90.8$103] 17.8$103 J/mol (Cho and Hwang, 2000)
ahydrolysis 3.20$1011 L/(mol s) [2.96$1011;3.44$1011]
bhydrolysis 2.40$108 s1 [2.65$108;2.15$108]
Ea,degradation 7.48$103 J/mol [6.90$103;8.06$103] 62.7$103 J/mol, (Cho and Hwang, 2000)
adegradation 2.30$103 L/(mol$s) [2.15$103;2.45$103]
bdegradation 1.80$104 s1 [1.65$104;1.95$104]
Ea,de-esterification 3.0$104 J/mol [1.38$104;4.62$104]
DE0 71.9% [71.6;72.2]
ade-esterification 11.0 L/(mol s) [10.8;11.2]
bde-esterification 5.01$102 s1 [4.80$102;5.22$102]
Ea,IV 5.32$104 J/mol [5.00$104;5.64$104]
IV0 0.778 L/g [0.755;0.801]
aIV 22.0$103 L/(mol s) [21.0$103;23.0$103]
bIV 0.793$103 s1 [0.780$103;0.807$103]
 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
Fig. 4. Pectin bulk solution concentration (mg/mL) as function of time (min) for ex-
tractions without acid at temperatures of 60
C, 70
C, and 80
C. pH was measured to
3.57 for these conditions.
factor,k0,IV, are in the order 103 s1. The chosen centering temper-
ature for estimations in the Arrhenius expressions is T0 ¼ 70
C as
the mean value in the composite design of experiments.
The initial guesses of the other model parameters are found in
literature and listed in Table 2 together with the estimated values of
the model parameters and 95% confidence intervals on the model
parameter. All relevant data has been used for estimation of each
subset of parameters to reduce the covariance and the confidence
limits of the estimates. The total surface area of 0.5 kg of peel is
calculated toATotal ¼ 0.48 m2.
4. Results
Simulations of pectin extractions from lime peel were per-
formed with the process parameters given in Table 1 and compared
to the data used in the parameter estimation to check how well the
Fig. 5. Pectin bulk solution concentration (mg/mL) as function of time (min) for ex-
tractions at pH ¼ 1.5 and temperatures of 60
C, 70
C, and 80
C.
67
model fit the observations. Fig. 4 shows simulated and measured
values of pectin bulk solution concentration for the extractions
without acid addition for 2 h at temperatures of 60, 70, and 80
C
which led to a pH of 3.75 during the extractions. The short
extraction time is used as this data will be part of the estimation of
some of the initial properties of the peel and the diffusive transport
only.
The order of magnitude for the diffusion coefficients calculated
from equation (6) with the estimated parameters given in Table 2 is
1011 m2/s, which is in agreement with the values found in (Minkov
et al., 1996). The transport of pectin by diffusion increases with
temperature. Fig. 4 shows that the highest pectin bulk solution
concentration is obtained at 80
C.
Figs. 5e7 show simulated and measured values of pectin bulk
solution concentration for extractions at pH values of 1.5, 2.3, 3.1
and at temperatures of 60, 70 and 80
C. The proposed model de-
scribes the pectin bulk solution concentration in the pilot scale
extractions to a satisfactory degree. The minor deviations between
simulations and measurements are due to statistical scattering of
the data.
The order of magnitude of the hydrolysis rate constant calcu-
lated from equation (6) using the estimated parameters in Table 2
varies depending on the specific temperature and pH. Values for
the hydrolysis rate constants in pectin extractions at different
temperatures and pH values have been reported previously (Paga n
and Ibarz, 1999). They modelled an overall hydrolysis rate constant
and therefore those results cannot be directly compared to the
values found in the present study. When the acid concentration is
very low, [Hþ] ¼ 7.5$104 M, the hydrolysis reaction rate given in
equation (6) approaches zero, meaning that there is not a sufficient
amount of catalyst in the solution to promote the hydrolysis of
protopectin.
The order of magnitude for the degradation rate constant
calculated from equation (3) using the estimated parameters in
Table 2 is 105 s1, which is in agreement with the parameter value
found in Cho and Hwang (2000). As expected the highest pectin
bulk solution concentration for a constant level of pH is obtained at
the highest extraction temperature. For an extraction temperature
of 80
C and pH ¼ 1.5 there is a significant decrease in pectin bulk
solution concentration which is reasonable since these conditions
also promote the degradation of pectin as described in Cho and
Hwang (2000). The reason that degradation leads to a reduced
observed pectin mass concentration is that the FIPA analysis does
not detect very short polymer chains. The degradation reaction
leads to a decrease in the chain length for the size distribution of
pectin molecules and a fraction therefore falls below the threshold
for detection.
Figs. 5e7 show that the higher pectin bulk solution concentra-
tions were obtained at the lower pH values, due to the catalytic
effect of the acid concentration. These results are consistent with
the conclusions of Paga n and Ibarz (1999) and Wai et al. (2010),
reporting that the effects of pH and extraction time seemed to be
more influential than temperature on pectin yield.
Figs. 8e10 display simulated and measured values of IV and
Figs. 11e13 show simulated and measured values of DE for ex-
tractions at pH values of 1.5, 2.3, and 3.1 and at temperatures of 60,
70 and 80
C. The proposed model describes IV and DE in the pilot
scale extractions to a satisfactory degree. The minor deviations
between simulations and measurements of IV and DE are due to
statistical scattering of the data.
A high IV means that the extracted pectin has a high gelling
quality (Pag an et al., 1999). In Figs. 8e10 it is shown that increased
temperature and acidity causes IV to decrease faster as also re-
ported by Paga n et al. (1999). In this study the highest IV was ob-
tained after 300 min by extraction at 60
C and pH ¼ 3.1. This
68 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
Fig. 6. Pectin bulk solution concentration (mg/mL) as function of time (min) for ex-
tractions at pH ¼ 2.3 and temperatures of 60
C, 70
C, and 80
C.
indicates that pectin with high IV and DE properties could be ob-
tained at low temperature and acidity but these combinations also
result in low pectin yield.
In Figs. 11e13 it is shown that increased temperature and
especially acidity cause DE to decrease faster as previously reported
by Wai et al. (2010). There is a large decrease in DE during ex-
tractions at pH ¼ 1.5 whereas only a small decrease in DE was
observed for extractions at pH ¼ 3.1. This shows that DE is mainly
influenced by pH as also reported by Wai et al. (2010).
5. Discussion
The model presented in equations (1) to (8) and (10) either
differs from or extend other first principle models published in the
n et al.,
literature (Panchev et al., 1989; Minkov et al., 1996; Dura
2015) by calculating not only the yield of extraction pectin but
Fig. 7. Pectin bulk solution concentration (mg/mL) as function of time (min) for ex-
tractions at pH ¼ 3.1 and temperatures of 60
C, 70
C, and 80
C.
Fig. 8. Intrinsic viscosity (dL/g) as function of time (min) for extractions at pH ¼ 1.5
and temperatures of 60
C, 70
C, and 80
C.
also the DE and IV, as well at taking the effects of temperature and
acid concentration on extraction kinetics into account. These other
reported models of pectin extraction are evaluated using data from
extractions in lab scale. The model proposed in this paper was
evaluated on pilot scale data, which resembles more closely the
extraction dynamics at an industrial scale. The present model was
developed as a general extraction model but so far only fitted based
on data from one type and source of peel. The same principles can
be used to model extraction of different types of fruit peels. In
general the quality of the model fits in relation to the data in
Figs. 4e13 was satisfactory. When evaluating the statistics, it is clear
that the band for the 95% confidence limits is quiet wide in some of
the parameter estimates. In general it is always desired to have
more observations to increase the certainty of the estimates which
would require either more samples or extra experiments. We do
however believe that based on the type of measurements we have
available for a pectin extraction process in either pilot or industrial
Fig. 9. Intrinsic viscosity (dL/g) as function of time (min) for extractions at pH ¼ 2.3
and temperatures of 60
C, 70
C, and 80
C.
 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
Fig. 10. Intrinsic viscosity (dL/g) as function of time (min) for extractions at pH ¼ 3.1
and temperatures of 60
C, 70
C, and 80
C.
scale the results are reasonable. Further improvements could
potentially be achieved if some of the general parameters could be
isolated and estimated in lab scale such some of the reaction ki-
netics phenomena and then the rest of the model parameters could
be tuned based on pilot or full scale data. We believe that the
approach taken to isolate model phenomena and estimating the
model parameters sequentially in as described in section 2.1 is the
best possible approach when all parameters are unknown.
While the focus in section 3 and 4 has been on variables that can
be measured in the bulk solution in the extraction tank such as
pectin yield, DE and IV that is not the only information a first
principle model of the extraction can provide. Since the reaction
and transport phenomena is modelled inside the peel, assuming
gradient in one dimension only, predictions of concentration pro-
files inside the peel can be computed. Fig. 14 shows such predicted
pectin concentration profiles at different special points, x inside the
Fig. 11. DE as function of time (min) for extractions at pH ¼ 1.5 and temperatures of
60
C, 70
C, and 80
C.
69
Fig. 12. DE as function of time (min) for extractions at pH ¼ 2.3 and temperatures of
60
C, 70
C, and 80
C.
peel using the estimated model parameters. It is seen as the pro-
topectin in the peel is hydrolysed the pectin concentration in the
peel is growing but due to the transport the concentrations at the
center of the geometry will assume the highest values. The con-
centration at the peel surface and in the bulk solution seems to be
on top of each other but a small gradient over the boundary layer
does exist. Hence it can be seen that the internal transport inside
the peel is slower that the transport over the boundary layer.
Further extractions at these conditions appear to be controlled by
the transport phenomena as the hydrolysis of protopectin is pre-
dicted to be fast in relation to the diffusion inside the peel. It is also
seen than the system converges towards a steady state where no
concentration gradients in the pectin concentrations is present in
the system, hence some amount of pectin will still be inside the
peel.
Extensions of the model to describing the change in viscosity in
the bulk solution could potentially strengthen the model. A
Fig. 13. DE as function of time (min) for extractions at pH ¼ 3.1 and temperatures of
60
C, 70
C, and 80
C.
70 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71

Fig. 14. Predicted pectin concentration profiles at different special points inside the peel as function of time for extractions at 70
C and pH 2.3.

different viscosity could affect diffusion and thereby the extraction Da Coefficient for to the energy resistance to diffusion [J/mol]
kinetics. Compared to industrial pectin extraction the present pilot DE: Degree of esterification [%]
plant extraction was performed with a low reasonable viscosity and Ea,hydrolysis Energy of activation for hydrolysis [J/mol]
it is therefore not reasonable to include viscosity effects in this Ea,degradation Energy of activation for degradation [J/mol]
study. Ea,deesterification Energy of activation for de-esterification [J/mol]
Ea,IV Constant for intrinsic viscosity [J/mol]
6. Conclusion fGA: Fraction of galacturonic acid
Hþ: Proton concentration [mol/L]
The proposed dynamic model in this work provided a mathe- L: Half thinness of peel slabs [m]
matical description of pectin extraction in terms of yield, degree of m: Mass [kg]
esterification and intrinsic viscosity. The dynamic model included IV: Intrinsic viscosity [L/g]
internal diffusion of pectin in the peel, reaction kinetics of hydro- k0,deesterification Arrhenius pre-exponential factor for de-
lysis, degradation, de-esterification and mass transfer of pectin esterification [s1]
from the solid phase to the bulk solution. The model includes the kdeesterification Reaction rate for the de-esterification[s1]
effects of temperature and acid concentration on extraction ki- k0,degradation Arrhenius pre-exponential factor for degradation [s1]
netics. A procedure for model calibration has been presented and it kdegradation Reaction rate for degradation [s1]
has been demonstrated that the model describes the pectin bulk k0,hydrolysis Arrhenius pre-exponential factor for hydrolysis [s1]
solution concentration, DE and intrinsic viscosity in pilot scale ex- khydrolysis Reaction rate for acidic hydrolysis [s1]
tractions from lime peel at different temperatures {60, 70, 80} de- kmasstransfer Mass transfer coefficient [m/s]
gree centigrade and pH {1.5, 2.3, 3.1} values. k0,IV Factor for intrinsic viscosity [s1]
kIV First order constant for intrinsic viscosity [s1]
Acknowledgement R Gas constant [J/(mol$K)]
T Temperature [K]
The authors would like to acknowledge the Technical University t Time [s]
of Denmark (DTU), Copenhagen University (KU) and BIOPRO2. The V Volume [m3]
project received financial support from Innovation Fund Denmark x Spatial coordinate within peel slabs [m]
through the BIOPRO2 strategic research center (Grant number a Slope in the linear model for the pre-exponential
4105-00020B). Arrhenius parameter [L/(mol$s)]
b Intercept in the linear model for the pre-exponential
Notation Arrhenius parameter [s1]
h Viscosity [kg/(s$m)]
A Surface area [m2]
cester(bulk) Concentration of ester groups on pectin in the bulk References
solution [kg/m3]
Cho, Y., Hwang, J., 2000. Modeling the yield and intrinsic viscosity of pectin in acidic
cpectin(peel) Concentration of pectin inside the peel [kg/m3] solubilization of apple pomace. J. Food Eng. 44, 85e89.
cpectin(bulk) Concentration of pectin in the bulk solution [kg/m3] n, R., Villa, A.L., Ribeiro, R., Rabi, J.A., 2015. Pectin extraction from mango peels
Dura
cprotopectin(peel) Concentration of protopectin [kg/m3] in batch reactor: dynamic one-dimensional modeling and lattice boltzmann
simulation. Chem. Prod. Process Model. 10, 203e210.
Dpectin Diffusion coefficient of pectin [m2/s] Genchev, L., Damyanov, D., Vladimirov, G., et al., 1988. Experimental determination
D0 Arrhenius pre-exponential factor for diffusion [m2/s] of the coefficient of diffusion of apple pectin. KhranitelnaPromishlenost. 37,
 N.M. Andersen et al. / Journal of Food Engineering 192 (2017) 61e71
24e26.
Girisuta, B., Janssen, L.P.B.M., Heeres, H.J., 2007. Kinetic study on the acid-catalyzed
hydrolysis of cellulose to levulinic acid. Ind. Eng. Chem. Res. 46, 1696e1708.
Graham, R.P., Sheperd, A.D., 1953. Pilot plant production of low-methoxyl pectin
from citrus peel. J. Agric. Food Chem. 1, 993e1001.
Himmelblau, D.M., 1970. Process Analysis by Statistical Methods. Wiley, New York.
Kanmani, P., Dhivya, E., Kumaresan, K., 2014. Extraction and analysis of pectin from
citrus peels: augmenting the yield from citrus limon using statistical experi-
mental design. Iran. J. Energy & Environ. 5, 303e312.
Kirtchev, N., Panchev, I., Kratchanov, C., 1989. Kinetics of acid-catalyzed de-esteri-
fication of pectin in a heterogeneous medium. Int. J. Food Sci. Technol. 24,
479e486.
Marra, F., 2016. Virtualization of processes in food engineering. Editor. e J. Food Eng.
176, 1.
Masmoudi, M., Besbes, S., Chaabouni, M., Robert, C., Paquot, M., Blecker, M., Attia, H.,
2008. Optimization of pectin extraction from lemon by-product with acidified
date juice using response surface methodology. Carbohydr. Polym. 74, 185e192.
Minkov, S., Minchev, A., Paev, K., 1996. Modelling of the hydrolysis and extraction of
apple pectin. J. Food Eng. 29, 107e113.
Pagan, J., Ibarz, A., 1999. Extraction and rheological properties of pectin from fresh
peach pomase. J. Food Eng. 39, 193e201.
Pagan, J., Ibarz, A., Llorca, M., Coll, L., 1999. Quality of industrial pectin extracted
from peach pomace at different pH and temperatures. J. Sci. Food Agric. 79,
1038e1042.
Panchev, I.N., Kirtchev, N.A., Kratchanov, C., 1989. Kinetic model of pectin extraction.
Carbohydr. Polym. 11, 193e204.
Saguy, I.S., 2016. Challenges and opportunities in food engineering: modeling, vir-
tualization, open innovation and social responsibility. J. Food Eng. 176, 2e8.
Savary, B.J., Hotchkiss, A.T., Fishman, M.L., Cameron, R.G., Shatter, R.G., 2003. In:
71
Voragen, F., Schols, H., Visser, R. (Eds.), Development of a Valencia Orange Pectin
Methyl Esterase for Generating Novel Pectin Products, Advances in Pectin and
Pectinase Research. Kluwer Academic Publishers, The Netherlands,
pp. 345e361.
Seader, J.D., Henley, E.J., 2006. Separation Process Principles, second ed. John Wiley
& Sons, New Jersey.
Shampine, L.F., Reichelt, M.W., 1997. The MATLAB ode suite. SIAM J. Sci. Comput. 18,
1e22.
Sirisakulwat, S., Nagel, A., Sruamsiri, P., Carle, R., Neidhart, S., 2008. Yield and quality
of pectins extractable from the peels of Thai mango cultivars depending on fruit
ripeness. J. Agric. Food Chem. 56, 10727e10738.
Sriamonsak, S., 2003. Chemistry of pectin and its pharmaceutical uses: a review.
Silpakorn Univ. Int. J. 3, 206e228.
Voragen, F., Schols, H., Visser, R.G.F., 2003. Advances in Pectin and Pectinase
Research. Springer.
Wai, W.W., Alkarkhi, A.F., Easa, A.M., 2009. Optimization of pectin extraction from
durian rind (Duriozibethinus) using response surface methodology. J. Food Sci.
74, C637eC641.
Wai, W.W., Alkarhi, A.F.M., Easa, A.M., 2010. Effects of extraction conditions on yield
and degree of esterification of durian rind pectin: an experimental design. Food
Bioprod. Process. 88, 209e214.
Willats, W.G.T., Knox, J.P., Mikkelsen, J.D., 2006. Pectin: new insights into an old
polymer are starting to gel. Trends Food Sci. Technol. 17, 97e104.
Winning, H., Viereck, N., Nørgaard, L., Larsen, J., Engelsen, S.B., 2007. Quantification
of the degree of blockiness in pectins using 1H NMR spectroscopy and che-
mometrics. Food Hydrocoll. 21, 256e266.
Zhou, W., Widmer, W., Grohmann, K., 2008. Developments in ethanol production
from citrus peel waster. Proc. Fla. State Hortic. Soc. 121, 307e310.