Renewable Energy 116 (2018) 857e865

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Biodiesel production in a continuous packed bed reactor with recycle:

A modeling approach for an esterification system

ndez-Montelongo a, Juan Paulo García-Sandoval a, *,
Rosaura Herna
Alejandro Gonza


lez-Alvarez a
n Aguilar-Garnica c, **
, Denis Dochain b, Efre
a
n, 44430, Guadalajara, Mexico
Departamento de Ingeniería Química, Universidad de Guadalajara, 1421 Blvd. Marcelino Garcia Barraga
b
ICTEAM, UCL, 4 Avenue Georges Lemaître, 1348, Louvain-la-Neuve, Belgium
c
Departamento de Química, Universidad Auto
noma de Guadalajara, 1201 Av. Patria, 44100, Guadalajara, Mexico

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

Article history: The aim of this work is to propose and to validate a dynamical model for an esterification reaction that
Received 17 March 2017 was carried out in a relatively new configuration for biodiesel production process: a packed bed reactor
Received in revised form with input, output, and recycle. More specifically, the reagents of the esterification process were free
26 July 2017
fatty acids (FFA) contained in canola oil and low cost methanol (LCM), whereas a commercial cation
Accepted 9 September 2017
Available online 18 September 2017
exchange resin (Purolite CT725) was used as a solid-acid catalyst. The effects of both the LCM volumetric
inflow rate and the inflow FFA molar fraction were investigated. The dynamical model was built
considering the following assumptions: 1) the reactor was modeled as a CSTR due to the high ratio of
Keywords:
Packed bed reactor
recycling volumetric flow to reactor volumetric inflow and 2) the kinetic expression in the model is
Esterification written in terms of activity coefficients in the light of the non-ideal behavior of the esterification reac-
Biodiesel production tion. A satisfactory agreement between the experimental data and the proposed model was obtained.
Heterogeneous catalyst Thus, the resulting model may be advised as a valuable tool for the analysis, control, and optimization of
Dynamical model continuous packed bed reactors where the esterification of FFA in the presence of solid acid catalyst is
carried out to produce biodiesel.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction homogeneous catalyst such as the sulfuric acid. However, these

Biodiesel is a renewable, biodegradable, and non-toxic blend of
esters that can replace diesel obtained from petroleum, implying a
significant emissions reduction [1]. It is usually produced from
refined oils in a transesterification process by homogeneous acid or
base catalysts [2]. However, the use of refined oils as raw materials
has a negative impact on the price of biodiesel. This is why low-cost
oils such as waste cooking oils, grease trap wastes or non-edible
soapstock have been alternatively used as raw materials [3].
Nevertheless, these low-cost oils are usually characterized for a
high content of Free Fatty Acids (FFA's) that are converted to soap if
the conventional transesterification process is directly applied [4].
Thus, these oils are converted into biodiesel in a two step (esteri-
fication-transesterification) process in which the first step or pre-
treatment (i.e. esterification) is usually catalyzed by an
* Corresponding author.
** Corresponding author.
E-mail addresses: paulo.garcia@cucei.udg.mx (J.P. García-Sandoval), efren.
aguilar@edu.uag.mx (E. Aguilar-Garnica).
catalysts generates large volumes of waste that impact both the
cost and environmental production of biodiesel [5]. An alternative
to eliminate these problems is to carry out the FFA esterification
process with solid catalysts, whose incorporation in the reaction
medium might also avoid corrosion, since the solid catalyst has
some water tolerance. Another benefit of using solid catalysts is
that they can be economically separated from products and re-used
[6,7]. The solid catalysts that are usually considered to produce
biodiesel from high FFA's oils in esterification processes are: mac-
roporous exchange resins [8], carbon supported catalysts [9], cat-
alysts derived from biomass such as lignin [10] and microporous
zeolites [11], only to mention a few examples.
Diverse mathematical models have been derived for FFA ester-
ification reactions catalyzed by solid catalysts. For instance, Tesser
et al. [12,13] have derived a Eley-Rideal type kinetic model for the
esterification of oleins and for a mixture formed by soybean oil with
oleic acid. The esterification reactions have been conducted in a
batch reactor containing the following acidic solid catalysts:
Amberlyst 15, Amberlyst 16, and Amberlyst 131. Another experi-
ment conducted in batch reactor is the one developed by

https://doi.org/10.1016/j.renene.2017.09.030
0960-1481/© 2017 Elsevier Ltd. All rights reserved.
858
ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna
Russbueldt and Hoelderich [14]. In this case, they have proposed a
pseudo-second order kinetic model for the esterification (in a batch
reactor) of various acidified oils in the presence of the solid acid
catalysts: EBD-100, EBD-200, and EBD-300. A pseudo-
homogeneous second order reversible model for the esterification
of the myristic acid with methanol in the presence of triglycerides
over sulfated zirconia in a batch reactor has been proposed by
Rattanaphra et al. [15]. Biodiesel can also be prepared in continuous
packed bed reactors as it was proved by Yun and Ling [16] in the
esterification of waste oils with the cation exchange resins: D002,
002CR and 733. Another continuous process of biodiesel produc-
tion is the one developed by Lu et al. [17], which consists of a fixed-
bed and a down-stream plug-flow reactors. This process was setup
to test different feedstock oils, and a type of ion-exchange resin was
used Lu et al. [17]. Even when there exists a wide variety of
mathematical models describing FFA esterification reactions cata-
lyzed by solid acids (either in batch or continuous processes), to the
best of our knowledge, mathematical models describing FFA
esterification reactions conducted within continuous packed bed
reactors with recycle have not been reported in the literature. Then,
the novelty of this contribution is to conduct a FFA esterification
process in this class of reactors and also to propose and to validate a
dynamical mathematical model describing such a process. Without
loss of generality, the dynamical model has been validated for the
esterification of a blend of oleic acid with canola oil in the presence
of low cost methanol which is being catalyzed by a cation exchange
resin Purolite CT725. For this task, we have considered the usual
sequence of modeling steps described in literature [18]: problem
definition, conceptual model, mathematical model, estimation of
parameters or calibration and validation. These steps are described
in section 3, while the results are described in section 4.
Fig. 1. Schematic view of the esterification reaction system.
2. Materials and methods
2.1. Materials
Oleic acid in the form of tallow fat was provided by QUIMIKAO
(Guadalajara, Jalisco, Me
xico). This fat was mixed with commercial
canola oil and the resulting acidified canola oil was used as raw
material in the esterification experiments. Cation exchange resin
Purolite CT2175 has been used as the solid catalyst. Purolite CT275
is a macro-reticular catalytic polymer, with a spherical bead phys-
ical appearance, and it is strongly acid. The catalytic resin Purolite
CT275 characterization, for instance, with scanning electron mi-
croscopy (SEM) and determination of its surface area, is extensively
reported in the literature [19,20].
The other reagent in these experiments was industrial reagent
grade methanol (Pochteca, Me
xico) characterized by a relatively
low moisture content (0.3 wt.%) and low cost, KOH, phenolptalein
(Golden Bell, USA) and isopropyl alcohol (Almace
n de drogas La Paz,
Me
xico) were used in the analysis of samples taken from the re-
action medium.
2.2. Experimental methods
2.2.1. Configuration of the esterification system in a packed bed
reactor with recycle
The esterification reaction was carried out in the system shown
in Fig. 1. In this system, R-1 is a 132 mL packed bed glass reactor (L
∕D ¼ 9). The packed bed reactor was immersed in a water bath (TT-
3) and its temperature was monitored with a J-type thermocouple
(TE) that was connected to a temperature controller (TC) (Love
controls, 16C series) that activated a relay (EY). The blend of tallow

ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna 859

fat and commercial canola oil (acid oil, Fao) was introduced to R-1
from a 50 mL stirred tank (TT-2) with a peristaltic pump (P-2)
(Control company, 3385 model), while industrial methanol (Fim)
was supplied to R-1 from a 250 mL stirred tank (TT-1) using a
dosing pump (P-1) (Grundfos, DDC 6e10 model). To avoid phase
separation in the inflow mixture, both streams were mixed very
closely to the input of the reactor. A couple of condensers (E1) and
(E2) have been connected to TT-1 and R-1 respectively, to avoid
methanol losses. A peristaltic pump (P-3) (Thermo Scientific
Manostat. 72-312-000 model) was used to recycle the reacting
mixture in R-1 in such a way that the ratio of recycling volumetric
flow to reactor volumetric inflow is high to approximate the
packed-bed reactor to a stirred-tank reactor. A stainless steel mesh
retention (RM) with 104 mm aperture is collected to keep the
catalyst in the reactor. Finally, the effluent (Fout) was captured in a
settler (FL-1), in which it was possible to separate the organic phase
(composed mainly by ester and unreacted acidified oils) from the
aqueous phase (composed mainly by water and methanol).
2.2.2. Operational conditions
Five experimental runs (ER1-ER5) were carried out at 58+ C.
Throughout the full experimental runs, the recycling volumetric
flow was such that the packed-bed reactor can be approximated to
a stirred-tank reactor. The reaction system was initially filled with
the catalyst (Purolite CT275), industrial methanol and the acidified
canola oil as described in Table 1. Then, it was operated for around
60 min in batch conditions, followed by the input-output (i.e.
continuous) operation. During the continuous operation, the
volumetric inflow rate of the acid oil remained constant while
diverse step changes were carried out in both the inflow FFA molar
fraction in the acidified oil and in the volumetric inflow rate of
industrial methanol, also described in Table 1. More specifically, an
increase in the volumetric inflow rate of industrial methanol has
been performed for the experimental runs ER1 and ER2, while
several step changes (including both increases and decreases) in
the volumetric inflow rate of industrial methanol were induced in
the experimental run ER5, whereas some increases were induced to
the inflow FFA molar fraction in the acidified oil during the ex-
periments ER3 and ER4. These variations in the volumetric inflow
rates and in the compositions are graphically depicted in Fig. 2.
Note that an abrupt variation at 220 min for the volumetric flow
rates of industrial methanol and acid was presented during the
experiment ER4. These variations were a consequence of the lack of
electricity in the facility where the experiment was conducted.
Nevertheless, this flaw was immediately fixed. Besides, the amount
of catalyst (Purolite CT275) initially packed within the reactor was
such that the packed bed length was 24 cm. Finally, it is important
to remark that for all these runs the recycle flow rate was
620 mL,min1 , that allowed to increase the contact area between
the immiscible oleic and aqueous phases, and represents a ratio of
recycling volumetric flow to reactor volumetric inflow around 850.
2.2.3. Sampling and analysis
The FFA's content in the organic phase obtained from the settler
located at the output of the reactor: u*F , was measured every hour
using the AOCS official method Ca 5a-40 [21]. This method implies
to dissolve the organic phase sample in isopropyl alcohol and to
titrate the dissolved sample with 0.1 N KOH solution using phe-
nolptalein as an indicator.
3. Dynamical modeling
3.1. Problem definition
The modeling goal in this contribution is to deduce and to
validate a mathematical model describing the dynamic behavior of
the compositions of FFA's, canola oil, methanol, esters and water in
the esterification reaction system previously described. In the
following sections the conceptual idea and the mathematical
modeling is presented.
3.2. Conceptual model
The esterification that has been conducted within the packed
bed reactor can be schematized by the general equation:
c
F þ A# E þ W (1)
where F represents the FFA's, A is the alcohol (in this case, meth-
anol), c is a solid acid catalyst, E denotes the esters, and W is water.
However, it is well known that the esterification is a complex
process composed of several steps and intermediate species with a
non-ideal behavior due to the presence of the acid catalyst, oily, and
aqueous compounds.
A kinetic model that considers the non-ideal behavior of an acid
catalyzed esterification of FFA's derived regarding the well-known
FischereSpeier esterification mechanism has been already pro-
posed by Herna
ndez-Montelongo et al. [22]. This kinetic expression
is based on activity coefficients instead of concentrations and it has
the following structure:
ac ðk1 aF aA  k2 aE aW Þ
rðaÞ ¼ ; (2)
2ðaF ; aA ; aE ; aW Þ
where ai ¼ giCi (i ¼ F,A,E,W,c) is the activity of the i-th component,
with gi and Ci as the activity coefficient and concentration of the

Table 1
Initial conditions and step changes for the experimental runs in the esterification reaction system.

Initial conditions Experimental run

ER1 ER2 ER3 ER4 ER5

Mass of Purolite CT275, mc (g) 41.68 41.29 41.29 41.20 41.66

Mass of acidified oil (g) 46.64 47.25 47.25 41.2 46.63
Mass fraction of FFA's in the acidified oil, wFao 0.528 0.516 0.516 0.513 0.528
Mass of industrial methanol (g) 55.91 55.39 55.39 55.26 55.90
Mass fraction of methanol in industrial methanol, wAim 0.997

Step change

Manipulated variable Fim Fim wFao wFao Fim,wFao

Type a a a a a,b
a
Increase.
b
Decrease, Fim industrial methanol volumetric flow.
860
ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna

acid. In Ref. [22] it was shown that, because some intermediates

react very fast, some of the terms in the denominator can be
neglected to obtain a kinetic expression with fewer kinetic pa-
rameters. In particular, for a homogeneous catalyst it was found
that the important terms were k10aFaW and k12aAaEaW. However, for
heterogeneous catalyst other terms could be the relevant ones.
Thus, the only difference between the kinetic model for esterifi-
cation catalyzed by a homogeneous catalyst and the kinetic model
(Eq. (2)) for a solid-acid catalyzed esterification will be the nu-
merical values of the kinetic parameters kj's (j ¼ 1; 2; …; 12) (For
more details about this fact, see section 3.4).
On the other hand, it is important to remark that the class of
reaction systems as the one under study, can be considered as a
intermediate case between a Plug Flow Reactor (PFR) and a
Continuous Stirred Tank Reactor (CSTR) [23]. If the recycle flow rate
in the reaction system is relatively low in comparison with the
inflow rate, then it can be considered as a PFR. In this case, the
mathematical model describing the behavior of the reagents within
the reaction system might be given by Partial Differential Equa-
tions. While, if the recycle flow rate is somewhat higher than the
inflow rate (e.g. > 25), then the reaction system is known as
“Recycle reactor”, i.e. a multi-pass reactor where a major portion of
the effluent is recycled to the reactor inlet, and can be viewed as a
CSTR [24], whose model is given by a set of Ordinary Differential
Equations (ODE's) that are derived from mass balances. As it was
explained in the previous section, the ratio of recycle flow rate and
inflow rate in all experimental runs was around 850, hence the
mathematical model describing the dynamic behavior of all the
chemical species involved in the esterification will be given by a set
of ODE's.

3.3. Mathematical model

The mass balances for each component of the esterification re-

action system under study are given in vectorial form by
Fig. 2. Inflow characteristics for the five experimental runs during continuous oper-
ation. (a) Industrial methanol volumetric inflow, (b) acidified oil volumetric inflow, and dn
(c) Inflow FFA molar fraction. ¼ q0  q þ dAc rðaÞ; (4)
dt

where t is time and n ¼ ½ nF nA nE nW nCoil T , with ni as the

same i-th component, respectively, and particularly, Cc is the con-
mole number of the i-th component (i ¼ F,A,E,W,Coil), being Coil a
centration of protons contained in the catalyst, that is considered
label selected to represent canola oil. In addition,
constant while the reaction is carried out. More specifically, this
T
concentration can be expressed as q0 ¼ qF;0 qA;0 qE;0 qW;0 qCoil;0 , where qi,0 are the molar
inflow rates of the i-th component such that
0 1
Cc ¼ bmc ; (3) rao Fao wFao
B MF C
B C
where mc is the catalyst mass of the resin within the reactor while b B C
0 1B r Fim wAim
C
is a constant to be determined. Besides, k1 and k2 are the forward qF;0 B im C
B C
and backward kinetic constants, while the denominator in this B qA;0 C B MA C
B C B C
kinetic expression, 2ðaF ; aA ; aE ; aW Þ ¼ 1 þ k3 aF þ k4 aA þ B qE;0 C ¼ B 0 C; (5)
B C B C
k5 aE þ k6 aW þ k 7 aF aA þ k8 aA aE þ k9 aE aW þ k10 aF aW þ k11 aF aA aW @ qW;0 A B C
B rim Fim ð1  wAim Þ C
þk12 aA aE aW , with parameters kk's (k ¼ 3; 4; …; 12), is the result of qCoil;0 B C
B MW C
considering a quasi-steady state in the intermediate reactions. This B C
B C
mechanism only considers the role of the proton's acid in the re- @ rao Fao ð1  wFao Þ A
action, therefore it is not necessary to specify if the acid catalyst is MCoil
homogeneous or heterogeneous (i.e. a solid acid) with the reaction
medium. On the other hand, due to the high ratio of recycling Here Fao and Fim as the volumetric inflow rates of the acidified oil
volumetric flow to reactor volumetric inflow, it is plausible to and industrial methanol with densities rao and rim, while wFao and
consider that, in spite of the use of a solid catalyst, the diffusion rate wAim are the mass fraction of FFA's in the acidified oil and the mass
is higher than the reaction rate, and the reaction mechanism is fraction of methanol in industrial methanol, respectively. Besides,
dominated by the kinetic expression given by Eq. (2). This is why in in this expression, Mi is the molar mass of the i-th component.
this contribution it is assumed that this model is also able to Furthermore, q ¼ ½ qF qA qE qW qCoil T , where qi are the molar
describe the kinetic of esterification reactions catalyzed by a solid- outflow rates of the i-th component such that

ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna 861

0 1
nF Fout dnE n Fout
B VL C ¼ E þ Ac rðaÞ; (11c)
B C dt V  Vc
B C
B nA Fout C
0 1 B C
qF B V C rim Fim ð1  wAim Þ
B L C dnW nW Fout
B qA C B C ¼  þ Ac rðaÞ; (11d)
B C B nE Fout C dt MW V  Vc
B qE C ¼ B C; (6)
B C B V C
@ qW A B L C
B C
qCoil B nW Fout C dnCoil rao Fao ð1  wFao Þ nCoil Fout
B C ¼  ; (11e)
B V C dt MCoil V  Vc
B L C
B C
@ nCoil Fout A
VL dV
¼ F0  Fout þ Dvr Ac rðaÞ; (11f)
dt
where VL is the liquid volume in the reactor. Besides,
Finally, the mass percentage of FFA's in the outflow organic
d ¼ ½ dF dA dE dW dCoil T , where di is the stoichiometric num- phase, can be computed from this model as:
ber of the i-th component. According to the kinetic expression
given by Eq. (1) these numbers are given by dF ¼ dA ¼ 1 and 100MF nF
dE ¼ dW ¼ 1. Since it is being assumed that the canola oil does not uF ¼ : (12)
MF nF þ ME nE þ MCoil nCoil
react in the esterification process, dCoil ¼ 0. Finally, Ac is the catalyst
superficial area that is proportional to the catalyst mass of the resin In this model the only unknown parameters are the kinetic
within the reactor, i.e. Ac ¼ amc where, a is physical characteristic of parameters kj's (i ¼ 1; 2; …; 12) and the catalyst coefficient b, that
the catalyst that can be considered as a known constant. In addi- are related to the reaction rate and its values will be estimated as it
tion, r(a) is the molar reaction rate per unit of the catalyst super- is described in the following section. The remaining parameters in
ficial area described by Eq. (2). the model were either taken from the literature or obtained directly
P from the configuration and operational conditions of the reaction
The liquid volume in the reactor is considered to be VL ¼ i Mri ni ,
i
system (see Tables 2 and 3). Another fact regarding the proposed
where ri is the density of the i-th component. Thus, taking into
model is that the dynamics of the triacylglycerol (TAG) contained in
account Eq. (4), the time derivative of VL is given by
the acidified oil is not being considered. Definitively, the TAG suf-
fers a transesterification to glycerol, but we are assuming that the
dVL
¼ F0  Fout þ Dvr Ac rðaÞ; (7) transesterification reaction rate of TAG is relatively slow if a solid
dt
acid catalyst (such as CT725 Purolite) is involved [16]. This is why
P Mi we are also neglecting the formation and dynamics of glycerol. This
where Dvr ¼ i ri di represents the molar volume change due to the
assumption is in line with the conclusion drawn by Canakci and
reaction. In Eq. (7), F0 is the total volumetric inflow rate that can be
Gerpen [4] stating that and acid-catalyzed transesterification is
computed by
much slower than that alkali-catalyzed.
F0 ¼ Fao þ Fim ; (8)
while Fout is the total volumetric outflow rate. On the other hand,
the catalyst also occupies a volume (Vc) that is considered constant, 3.4. Estimation of parameters or model calibration
thus dVc∕dt ¼ 0.
Once the expressions for VL and Vc and their derivatives have Nonlinear regressions, considering experimental data of ER1,
been defined, the following expressions for the total volume of ER3 and ER5, have been carried out to estimate the numerical
reaction medium V and its derivative are straightforwardly values of the parameters ki's (i ¼ 1; 2; …; 12) and b that are con-
computed tained in the model given by Eqs. (11) and (12). Once the param-
eters have been computed, the magnitude of each term of the
V ¼ VL þ Vc ; (9) denominator of the kinetic expression given by Eq. (2),
rðaF ; aA ; aE ; aW Þ, was analyzed; this analysis showed that only one
dV kinetic parameter of the denominator is relevant (k12) reducing to
¼ F0  Fout þ Dvr Ac rðaÞ: (10) three the kinetic parameters for the model, namely k1, k2 and k12,
dt
thus the final kinetic expression used in this work is
It is important to remark that during the batch operation of the
process, Fout ¼ 0, but once the volume of the packed bed reactor
plus the volume of the resin (Vmax ) is reached, i.e. V ¼ Vmax , it Table 2
Properties of the chemical species involved in the esterification reaction system.
follows that Fout ¼ F0 þ Dvr Ac r(a) because such total volume is
constant. Compound Molecular Molecular Density,ri Stoichiometric
Summarizing, if one replaces Eqs. (5) and (6) into Eq. (4) and formula mass, Mi (g/mL) number, di

including Eq. (10) the model to be considered is Oleic acid (F) C18H34O2 282 0.8769a 1
Methanol (A) CH3OH 32 0.7550b 1
dnF rao Fao wFao nF Fout Methyl oleate (E) C19H36O2 296 0.8451c 1
¼   Ac rðaÞ; (11a) Water (W) H2O 18 0.9831d 1
dt MF V  Vc Canola oil (Coil) C57H102O6 882 0.892e 0
a
Density measured in laboratory to 60+ C.
dnA rim Fim wAim n Fout b
See reference [25].
¼  A  Ac rðaÞ; (11b)
dt MM V  Vc c
See reference [26].
d
See reference [27] and.
e
See reference [28].
862
ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna

Table 3
Operational parameters and properties used in the esterification reaction system.

Parameter Value

Temperature (K) 333.15

Total reactor volume including 164
volume of the resin, Vmax (mL)
Dvr (mL/mol) 7.5
Resin area per unit of massa, a (m2/g) 45
a
From Purolite CT275 data sheet.

ac ðk1 aF aA  k2 aE aW Þ
rðaÞ ¼
1 þ k12 aA aE aW
gc bmc
ðk1 gF nF gA nA  k2 gE nE gW nW Þ
ðV  Vc Þ2
¼
k12 gA nA gE nE gW nW
1þ
ðV  Vc Þ3
Both regressions (using kinetic expression Eq. (2) and kinetic
expression Eq. (13)) have been conducted by applying the Nelder-
Mead simplex algorithm [29] to find a minimum of an uncon-
strained function. In this contribution, such a function is a weighted
sum of squared residual defined to be
X  * T  * 
F¼ Wq !
u F;q  !
u F;q !
u F;q  !
u F;q ;
q
h iT
where /u*F;q ¼ u*F;q ð0Þ u*F;q ðt1;q Þ u*F;q ðt2;q Þ / u*F;q ðtn;q Þ

T
and /uF;q ¼ uF;q ð0Þ uF;q ðt1;q Þ uF;q ðt2;q Þ / uF;q ðtn;q Þ ,
with q as a subindex that is considered to label an experimental
run. Thus, q ¼ ER1, ER3 or ER5. Furthermore, u*F;q ðtk;q Þ is the
experimental result that, according to the sampling and analysis
subsection, has been recorded using the AOCS official method Ca
5a-40 at each q-th experimental run for the mass percentage of
FFA's measured from the organic phase obtained from the settler at
each instant tk;q ¼ 0; t1;q ; t2;q ; …; tn;q . Besides, uF,q(tk,q) is the mass
percentage of FFA's in the outflow organic phase at tk,q, obtained
with Eq. (12) when the proposed dynamical model Eq. (11a) is
solved. More specifically, this model has been solved with an
implementation of the numerical method TR-BDF2, an implicit
Runge-Kutta formula with a first stage that is a trapezoidal rule step
and a second stage that is a backward differentiation formula of
order two [30,31]. In addition, Wq are weighting factors computed
as Wq ¼ 1∕Nq, where Nq is the number of data available for the q-th
experimental run. It is very important to remark that the activity
coefficients, gi, used in the kinetic expression Eq. (13), have been
computed with the UNIFAC group-contribution method, the group
surface areas and volume contributions along with the interaction
parameters are depicted in the Appendix (see Ref. [22] for further
details).
3.5. Model validation
Once the parameters k1, k2, k12, and b have been estimated using
experiments ER1, ER3 and ER5, they were used to solve the pro-
posed dynamical model Eq. (11) using the conditions of the
experimental runs ER2 and ER4. The results obtained in this form,
have been compared with the experimental results obtained in ER2
and ER4, that are different/independent from those experimental
runs that have been considered to estimate the parameters. Then,
the model results and the experimental results have been analyzed
(13)
Fig. 3. Percentage of FFA's in the outflow organic phase for experimental runs ER1, ER3
and ER5 used to calibrate the proposed model. Experimental data in symbols and
model predictions in continuous lines.
to verify the validity of the proposed model. More specifically, the
F-test [32] with a significance level of 0.05 has been used.
(14) 4. Results and discussion
The experimental results of the continuous operation, obtained
by applying the procedure described in section 2.2 and the volu-
metric inflow rates depicted in Fig. 2, are shown in Figs. 3e5; Figs. 3
and 4 show the percentage of FFA's in the outflow organic phase
(u*F ), while Fig. 5 depicts the total volumetric outflow (Fout). In these
figures, symbols represent the experimental data and the contin-
uous lines are the predictions of model Eq. (11). Finally, Fig. 6 de-
picts the estimated activity coefficients.
Regarding Figs. 3 and 4 it is important to make the following
statements. First, in the five experimental runs (ER1-ER5), the
initial percentage of FFA's in the outflow organic phase is relatively
low. This is due to the batch operation (carried out in approximately
60min) that allow to attain the reaction equilibrium. Then, when
the continuous operation started, the percentage of FFA's in the
outflow organic phase shows an increase and the equilibrium is
reached at 400 min, approximately. Now, let us analyze some
particular results starting for runs ER1 and ER2, where increments
in the volumetric inflow rate of industrial methanol were induced
at 420 and 600 min, respectively. These increments might produce
two opposite effects: (1) an increase in the methanol concentration
within the reactor that produces the shift of the reaction equilib-
rium towards the products, decreasing the FFA concentration, and
(2) a decrease of the residence time, increasing the FFA concen-
tration. In the particular operating conditions for ER1 and ER2,
although the volumetric inflow rate of industrial methanol was
increased to more than twice, the percentage of FFA's in the outflow
organic phase decreased, indicating that the effect (1) is more
important.
On the other hand, the increments of the FFA's content in the
acidified oil during experimental runs ER3 and ER4, are the cause of
decrement in the Methanol:FFA's molar ratio. At the same time, this
was the cause of an increase in the percentage of FFA's in the
outflow organic phase due to a slower reaction rate. This phe-
nomenon is in agreement with previous works see for instance
[33]. Finally, the relatively small step changes (both increments and
decrements) in the volumetric inflow rate of industrial methanol
performed during experimental run ER5 cause an oscillation of

ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna 863

Fig. 4. Percentage of FFA's in the outflow organic phase for experimental runs ER2 and
ER4 used to validate the proposed model. Experimental data in symbols and model
predictions in continuous lines.

Fig. 5. Total volumetric outflow. Experimental data in symbols and model predictions Fig. 6. Estimated activity coefficients for the 5 experimental runs.
in continuous lines.

Table 4
around 12% for the percentage of FFA's in the outflow organic Estimated kinetic parameters.
phase.
Parameter Value
As it was stated in section (3.5), with the experimental data
2
obtained from ER1, ER3 and ER5, the parameters k1, k2, k12, and b
2
k1 (mL ,mol ,min1 ) 3:39  105
k2 (mL2 ,mol2 ,min1 ) 2:88  105
(whose numerical values are shown in Table 4) were estimated (i.e.,
k12 (mL3 ,mol3 ) 9:26  107
the model was calibrated). These estimated parameters, together
b (mol,mL1 ,g1 ) 1:29  103
with the parameters of the process reported in Table 2, the oper-
ational conditions depicted in Table 3, the inflow conditions re-
ported in Fig. 2 and the initial conditions reported in Table 1, were
used to simulate the dynamical behavior of of the esterification model are adequate, namely, (1) due to the high ratio of recycling
system for ER1 to ER5 which is given by Eq. (11). Fig. 3 shows the volumetric flow to reactor volumetric inflow the reactor could be
simulations for experimental runs ER1, ER3 and ER5 while Fig. 4 considered a CSTR and (2) in spite of the use of a solid catalyst, the
shows the predictions of the independent experimental runs ER2 diffusion rate might be higher than the reaction rate, and as a
and ER4 used to validate the model. These simulations correspond consequence, the reaction mechanism might be governed by the
only to the continuous operation of the esterification system kinetic expression given in Eq. (13).
although these simulations were started with the batch period, Due to the non-ideal behavior of the reacting mixture, as sug-
which is not shown because of the lack of experimental data during gested elsewhere [22], the proton activity plays a key role during
this period. In both figures the p-value for the significance test is the esterification reaction and therefore the kinetic expression Eq.
also reported. As it can be seen, both sets of data are satisfactorily (13) considers activities, instead of concentrations, which are
reproduced with a p-value smaller that the significance level typically used to describe homogeneous acid catalyzed esterifica-
p ¼ 0.05. Then, the corresponding correlation between experi- tion reactions. This conjecture can be confirmed analyzing Fig. 6
mental data and model predictions can be considered to be sig- that describes the behavior of the calculated activity coefficients
nificant. This suggests that the initial hypotheses to develop the for reagents and products during the development of the FFA
864
ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna

esterification. Notice that the activity coefficients for the water (gW) Acknowledgments
and the protons of the acid catalyst (gHþ ) at the beginning of the
batch stage vary considerably, in particular, gHþ decreases abruptly The author gratefully acknowledge financial support from the
(one order of magnitude approximately) once the water is pro- Consejo Nacional de Ciencia y Tecnología (CONACyT) through the
duced, whereas gW has an opposite effect to that of gHþ . After these project CB2009-133554 to perform this study.
abrupt variations, gW approaches to 1, while gHþ remains approx-
imately around 0.04. In particular, for experimental runs ER1 and
Nomenclature
ER2, gW decreases and gHþ increases approximately a half order of
magnitude after the step in the methanol inflow rate was carried
a Catalyst coefficient
out, while for experimental runs ER3 and ER4, gW and gHþ slightly
b Catalyst superficial area
increases and decreases, respectively, after the FFA molar fraction is
Dvr Molar volume change
increased. The changes in gW and gHþ produce a double effect on
d Stoichiometric number
the FFA consumption rate: first, the total reaction rate decreases
g Activity coefficients
due to the diminution on the acid activity coefficient, and second,
u Mass percentage
the water activity increases the reverse reaction rate. As water is
F Unconstrained function to be optimized
formed, it competes with the alcohol for the hydrogen ion, cutting
r Density
down the concentration of alcohol complexes, and hence slowing
a Activity
up the esterification. On the other hand, the activity coefficients for
Ac Resin area per unit of mass
esters (gE) and canola oil (gCoil) keep of the order of magnitude 101,
C Concentration
while for FFA's (gF) and alcohol (gA) remain around 1. The magni-
F Volumetric flow
tudes and variations on the activity coefficients imply that the non-
ki i-th kinetic coefficient
ideal behavior of the reaction should be considered and suggests
M Molar mass
that the acid catalyst activity has a key role in the reaction rate,
m Mass
since it can change even in orders of magnitude when few amounts
n Moles number
of water are produced or initially present, suggesting that the ef-
Nq Number of data available for the q-th experimental run
forts to optimize the FFA esterification reaction systems should be
qi;0 Molar inflow rate for the i-th component
focused in the design of processes where the water can be
constantly eliminated in order to avoid the drop of the acid catalyst qi Molar outflow rate for the i-th component
r Molar reaction rate per unit of catalyst superficial area
activity. On the other hand, it was found that the important terms in
the denominator of this kinetic expression for the esterification V Volume of reaction medium
Vc Volume of catalyst in the reaction medium
reaction with a heterogeneous catalyst are 1 and k12aAaEaW, in
contrast to the terms k10aFaW and k12aAaEaW, that were the impor- VL Volume of liquid in the reaction medium
Vmax Volume of packed bed reactor
tant ones for an esterification reaction with a homogenous catalyst
[22]. Notice that the denominator of Eq. (13) has the typical w Mass fraction
Wq Weighting factor for the q-th experimental run
structure of solid catalyzed kinetics like Langmuir or Langmuir-
Hinshelwood. A Alcohol
The surface area of the catalyst used in this study is very low, ao Acid oil mixture
however we decided to use a commercial cation exchange resin c Catalyst
(Purolite CT725) and industrial methanol to analyze the feasibility Coil Canola oil
of the esterification in industrial applications using common and E Esters
cheap materials. In addition, the resin Purolite CT275 is a macro- F Free fatty acids
reticular catalytic polymer, with a spherical bead physical appear- im Industrial methanol mixture
ance, and a particle diameter of approximately 1 mm, that allows an out Outflow property
easy separation with a stainless steel mesh retention (RM) used to W Water
keep the catalyst in the reactor and to avoid it to be dragged out the 0 Inflow property
reactor in the recirculation or outflow streams. Any other catalyst
with bigger surface area see for instance [34,35] and similar particle A UNIFAC group parameters
diameter could be used to improve the conversion rate, however
the reactor configuration and the modeling procedure presented in
this work would be still valid.

Table A.1
UNIFAC group volume Rk and surface area Qk parameters. Parameters for Hþ, SO2
5. Conclusions 4
and SO1
3 are from Ref. [36]. The rest are from Ref. [37].

A model for the esterification of FFA's in the presence of solid Main group Rk Qk

acid catalyst has been developed, calibrated and validated. The COOH 1.3013 1.224
esterification runs have been performed in a packed bed reactor CH2 0.6744 0.54
CH3 0.9011 0.848
with a recycling volumetric flow rate that allows to model the
CH]CH 1.1167 0.867
reactor as a CSTR. In addition, the model predictions are significant CH3OH 1.4311 1.432
and they suggest that the reaction process is dominated by the CH2COO 1.6764 1.42
chemical reaction and not by the diffusion of reactants towards the H2O 0.92 1.4
solid catalyst. Thus, the model may be used to further analyze the Hþ 1 1

dynamical behavior of the biodiesel production in packed bed re- SO1

3
1.64 1.6
CH 0.4469 0.228
actors with recycle for optimization and control purposes.

ndez-Montelongo et al. / Renewable Energy 116 (2018) 857e865
R. Herna 865

Table A.2
UNIFAC Group-Group interaction parameters, aij (K). Parameters for Hþ and SO1
3 are from Ref. [36]. The rest are from Ref. [37]. Some parameters are considered to be zero since
there is not experimental values.

i J

COOH CH2 CH3 CH]CH CH3OH CH2COO H2O Hþ SO1 CH

3

COOH 0 315.3 315.3 1264 1020 256.3 66.17 0 1970 315.3

CH2 663.5 0 0 86.02 697.2 232.1 1318 0 640.5 0
CH3 663.5 0 0 86.02 697.2 232.1 1318 0 640.5 0
CH]CH 318.9 35.36 35.36 0 787.6 37.85 270.6 0 0 35.36
CH3OH 289.5 16.51 16.51 12.52 0 10.72 181 0 0 16.51
CH2COO 660.2 114.8 114.8 132.1 249.6 0 200.8 0 0 114.8
H2O 14.09 300 300 496.1 289.6 72.87 0 1998 1818 300.0
Hþ 0 0 0 0 0 0 735.4 0 0 0
SO1 526.6 21.18 21.18 0 0 0 1620 0 0 21.18
3
CH 663.5 0 0 86.02 697.2 232.1 1318.0 0 640.5 0

References [18] A. Rasmuson, B. Andersson, L. Olsson, R. Andersson, Mathematical Modeling in

Chemical Engineering, Cambridge, first. edn, 2014.
[19] D. Patel, B. Saha, Heterogeneous kinetics and residue curve map (RCM)
[1] M. Ahmad, M. Khan, M. Zafar, S. Sultana, Practical Handbook on Biodiesel
determination for synthesis of n-hexyl acetate using ion-exchange resins as
Production and Properties, CRC Press, 2013.
catalysts, Ind. Eng. Chem. Res. 46 (10) (2007) 3157e3169, https://doi.org/
[2] S. Semwal, A.K. Arora, R.P. Badoni, D.K. Tuli, Biodiesel production using het-
10.1021/ie060725x.
erogeneous catalysts, Bioresour. Technol. ISSN: 0960-8524 102 (2011)
[20] M. Umar, D. Patel, B. Saha, Kinetic studies of liquid phase ethyl tert-butyl ether
2151e2161, https://doi.org/10.1016/j.biortech.2010.10.080.
(ETBE) synthesis using macroporous and gelular ion exchange resin catalysts,
[3] L. Lin, Z. Cunshan, S. Vittayapadung, S. Xiangqian, D. Mingdong, Opportunities
Chem. Eng. Sci. 64 (21) (2009) 4424e4432, https://doi.org/10.1016/
and challenges for biodiesel fuel, Appl. Energy. ISSN: 0306-2619 88 (4) (2011)
j.ces.2009.07.015.
1020e1031, https://doi.org/10.1016/j.apenergy.2010.09.029.
[21] I. Rukunudin, P. White, C. Bern, T. Bailey, A modified method for determining
[4] M. Canakci, J.V. Gerpen, Biodiesel production from oils and fats with high free
free fatty acids from small soybean oil sample sizes, J. Am. Oil Chemists Soc. 75
fatty acids, Trans. ASAE. ISSN: 2151-0059 44 (6) (2001) 1429e1439, https://
(5) (1998) 563e568, https://doi.org/10.1007/s11746-998-0066-z.
doi.org/10.13031/2013.7010.
[22] R. Herna
ndez-Montelongo, J.P. García-Sandoval, E. Aguilar-Garnica, On the
[5] J.R. Kastner, J. Miller, D.P. Geller, J. Locklin, L.H. Keith, T. Johnson, Catalytic
non-ideal behavior of the homogeneous esterification reaction: a kinetic
esterification of fatty acids using solid acid catalysts generated from biochar
model based on activity coefficients, React. Kinet. Mech. Catal. ISSN: 1878-
and activated carbon, Catal. Today. ISSN: 0920-5861 190 (1) (2012) 122e132,
5204 115 (2) (2015) 401e419, https://doi.org/10.1007/s11144-015-0848-x.
https://doi.org/10.1016/j.cattod.2012.02.006.
[23] H. Fogler, Elements of chemical reaction engineering, Prentice-Hall Int.
[6] J.A. Melero, J. Iglesias, G. Morales, Heterogeneous acid catalysts for biodiesel
(1999).
production: current status and future challenges, Green Chem. ISSN: 1463-
[24] J. Berty, Testing commercial catalysts in recycle reactors, Catal. Rev. sci. Eng.
9270 11 (9) (2009) 1285e1308, https://doi.org/10.1039/b902086a.
20 (1) (1979) 75e96, https://doi.org/10.1080/03602457908065106.
[7] Z. Wen, X. Yu, S.-T. Tu, J. Yan, E. Dahlquist, Synthesis of biodiesel from vege-
[25] S. Mikhail, W. Kimel, Densities and viscosities of methanol-water mixtures,
table oil with methanol catalyzed by Li-doped magnesium oxide catalysts,
J. Chem. Eng. Data 6 (4) (1961) 533e537, https://doi.org/10.1021/
Appl. Energy. ISSN: 0306-2619 87 (3) (2010) 743e748, https://doi.org/
je60011a015.
10.1016/j.apenergy.2009.09.013.
[26] M.J. Pratas, S. Freitas, M.B. Oliveira, S.C. Monteiro, A.S. Lima, J. Coutinho,
[8] J. Fu, L. Chen, P. Lv, L. Yang, Z. Yuan, Free fatty acids esterification for biodiesel
Densities and viscosities of fatty acid methyl and ethyl esters, J. Chem. Eng.
production using self-synthesized macroporous cation exchange resin as solid
Data 55 (9) (2010) 3983e3990, https://doi.org/10.1021/je100042c.
acid catalyst, Fuel. ISSN: 0016-2361 154 (2015) 1e8, https://doi.org/10.1016/
[27] J.G. Speight, Lange's Handbook of Chemistry, vol. 1, McGraw-Hill, New York,
j.fuel.2015.03.048.
2005.
[9] M. Toda, A. Takagaki, M. Okamura, J.N. Kondo, S. Hayashi, K. Domen, M. Hara,
[28] B. Esteban, J.-R. Riba, G. Baquero, A. Rius, R. Puig, Temperature dependence of
Green chemistry: biodiesel made with sugar catalyst, Nature. ISSN: 1476-4679
density and viscosity of vegetable oils, Biomass Bioenergy 42 (2012) 164e171,
438 (7065) (2005) 178, https://doi.org/10.1038/438178a.
https://doi.org/10.1016/j.biombioe.2012.03.007.
[10] F. Guo, Z.-L. Xiu, Z.-X. Liang, Synthesis of biodiesel from acidified soybean
[29] J. Lagarias, J.A. Reeds, M.H. Wright, P.E. Wright, Convergence properties of the
soapstock using a lignin-derived carbonaceous catalyst, Appl. Energy. ISSN:
nelder-mead simplex method in low dimensions, SIAM J. Optim. 9 (1) (1998)
0306-2619 98 (2012) 47e52, https://doi.org/10.1016/j.apenergy.2012.02.071.
112e147, https://doi.org/10.1137/S1052623496303470.
[11] A.M. Doyle, T.M. Albayati, A.S. Abbas, Z.T. Alismaeel, Biodiesel production by
[30] M. Hosea, L. Shampine, Analysis and implementation of TR-BDF2, Appl.
esterification of oleic acid over zeolite Y prepared from kaolin, Renew. Energy
Numer. Math. 20 (1e2) (1996) 21e37, https://doi.org/10.1016/0168-9274(95)
97 (2016) 19e23, https://doi.org/10.1016/j.renene.2016.05.067.
00115-8.
[12] R. Tesser, L. Casale, D. Verde, M. Di Serio, E. Santacesaria, Kinetics of free fatty
[31] R. Bank, W. Coughran, W. Fichtner, E. Grosse, D. Rose, R. Smith, Transient
acids esterification: batch and loop reactor modeling, Chem. Eng. J. ISSN:
simulation of silicon devices and circuits, IEEE Trans. Computer-Aided Des.
1385-8947 154 (1e3) (2009) 25e33, https://doi.org/10.1016/
Integr. Circuits Syst. 4 (4) (1985) 436e451, https://doi.org/10.1109/
j.cej.2009.03.010.
tcad.1985.1270142.
[13] R. Tesser, L. Casale, D. Verde, M. Di Serio, E. Santacesaria, Kinetics and
[32] A. Constantinides, N. Mostoufi, Numerical Methods for Chemical Engineers
modeling of fatty acids esterification on acid exchange resins, Chem. Eng. J.
with MATLAB Applications, Prentice Hall, 1999.
ISSN: 1385-8947 157 (2e3) (2010) 539e550, https://doi.org/10.1016/
[33] M. Canakci, J.V. Gerpen, Biodiesel production via acid catalysis, Trans. ASAE 42
j.cej.2009.12.050.
(5) (1999) 1203e1210, https://doi.org/10.13031/2013.13285.
[14] B.M. Russbueldt, W.F. Hoelderich, New sulfonic acid ion-exchange resins for
[34] T.M. Albayati, A.M. Doyle, Encapsulated heterogeneous base catalysts onto
the preesterification of different oils and fats with high content of free fatty
SBA-15 nanoporous material as highly active catalysts in the trans-
acids, Appl. Catal. A General. ISSN: 0926-860X 362 (1e2) (2009) 47e57,
esterification of sunflower oil to biodiesel, J. Nanoparticle Res. 17 (2) (2015)
https://doi.org/10.1016/j.apcata.2009.04.019.
109, https://doi.org/10.1007/s11051-015-2924-6.
[15] D. Rattanaphra, A.P. Harvey, A. Thanapimmetha, P. Srinophakun, Kinetic of
[35] A.M. Doyle, Z.T. Alismaeel, T.M. Albayati, A.S. Abbas, High purity FAU-type
myristic acid esterification with methanol in the presence of triglycerides over
zeolite catalysts from shale rock for biodiesel production, Fuel 199 (2017)
sulfated zirconia, Renew. Energy 36 (10) (2011) 2679e2686, https://doi.org/
394e402, https://doi.org/10.1016/j.fuel.2017.02.098.
10.1016/j.renene.2011.02.018.
[36] T. Raatikainen, A. Laaksonen, Application of several activity coefficient models
[16] L. Yun, W. Ling, Biodiesel preparation from waste oil using cation exchange
to water-organic-electrolyte aerosols of atmospheric interest, Atmos. Chem.
resin as heterogeneous catalyst, Chem. Technol. Fuels Oils 45 (6) (2009)
Phys. 5 (9) (2005) 2475e2495, https://doi.org/10.5194/acp-5-2475-2005.
417e424, https://doi.org/10.1007/s10553-010-0167-x.
[37] R.C. Reid, J.M. Prausnitz, T.K. Sherwood, The Properties of Gases and Liquids,
[17] P. Lu, Z. Yuan, L. Li, Z. Wang, W. Luo, Biodiesel from different oil using fixed-
1977.
bed and plug-flow reactors, Renew. Energy 35 (1) (2010) 283e287, https://
doi.org/10.1016/j.renene.2009.07.011.