You are on page 1of 11

Fuel Processing Technology 142 (2016) 326–336

Contents lists available at ScienceDirect

Fuel Processing Technology

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

Research article

Transesterification of oil to biodiesel in a continuous tubular reactor with


static mixers: Modelling reaction kinetics, mass transfer, scale-up and
optimization considering fatty acid composition
Blaž Likozar ⁎, Andrej Pohar, Janez Levec
Laboratory of Catalysis and Chemical Reaction Engineering, National Institute of Chemistry, Hajdrihova 19, 1000 Ljubljana, Slovenia

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

Article history: Simulations of the methanolysis of lipids in a mixer-packed bed heterogeneous flow process were performed
Received 29 July 2015 based on the thermodynamics (equilibrium) as well as mechanism and chemical kinetics of glycerides and
Received in revised form 7 October 2015 fatty acid methyl esters (FAME) containing different combinations of gadoleic (G), linoleic (L), linolenic (Ln),
Accepted 24 October 2015
oleic (O), palmitic (P) and stearic (S) acids, bonded as glycerol (alcohol) substituents. Transport phenomena
Available online xxxx
and fluid dynamics were established for emulsion interface films and through axial dispersion coefficients (Péclet
Keywords:
number) in a fixed bed for both transient and stationary operation. Alcoholysis was also studied experimentally
Biodiesel production process optimization within a broad range of temperatures, hydrodynamic conditions in terms of volumetric flow rates, phase ratios
Reaction mechanism and kinetics and base catalyst concentrations. Diffusion resistance proved to be negligible, temperature was proven as a prev-
Mass transfer and diffusion alent factor, while steady state was reached only after a few residence times elapsed. Process economics were fi-
Mathematical model nally evaluated in terms of cost and price breakdown, allowing methodology extension to other waste oil
Waste oil transesterification resources. Sensitivity analysis revealed high LLL triglyceride content in oil having the greatest positive, and
Renewable energy and fuels OOL negative effect on FAME yield. Process intensification with static mixers was proven feasible upon scaling-
up from 5 to 10 mm internal diameter with no deterioration in overall process progression.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction dissolved catalysts [1,2,5,6] are more widespread and economical, and
even more so, the ones using alkali catalysts [1,2,7–16] due to extremely
Petroleum-based energy in terms of fossil fuels to date accounts for high reaction rates. However, the mass transfer limitations pertinent to
the vast majority of the primary energy consumed in the world, and a biodiesel production call for an innovative reactor design and the opti-
predominant portion of that amount is used in the transportation sec- mization of the existing ones.
tor. The current exploration of alternative energy is driven by the need Consequently, various types of reactors were either designed or
to develop sustainable energy resources that are cheaper, renewable employed in order to optimize and intensify the continuous production
and environmentally friendly [1]. The majority of the commercially of biodiesel, for example reactors exploiting cavitation [2], spinning
available biodiesel, one of the most promising biofuels, is made by the tube reactors [2], microwave reactors [2], membrane reactors [2,7,12],
trans-esterification of vegetable oils and animal fats with methanol or reactors with oscillatory flow [2,8], high pressure reactors [7], micro-
ethanol in stirred-tank reactors in the presence of base or acid catalysts structured reactors [7,11,14], fibre reactors [9], film reactors [10], chan-
[2]; nonetheless, these reactors are often used for batch production, cre- nel reactors [15], and reactors with metal foam [16]. These reactors pose
ating the demand for continuous reactors and processes, and their cor- an alternative to continuous stirred-tank reactors, which are the most
responding scale-up, optimization and intensification. widely utilized type in industrial practise, but exhibit several disadvan-
The continuous production of biodiesel may classified by the type of tages [2,7].
the catalyst used in the process. Recently, a few studies presented the All of these reactor configurations without a doubt optimize and in-
continuous production without any catalyst, either at relatively high tensify the biodiesel production; nonetheless, most of them are related
temperatures [3], or at supercritical conditions [4]. The trans- either to high investment costs and low overall production capacity (i.e.
esterification using heterogeneous or enzymatic catalysis draws more micro-structured and channel reactors), or high operating costs and
and more attention; nevertheless, the processes involving acidic challenging process' safety regulations (i.e. high pressure and micro-
wave reactors). A simple and robust design of tubular reactors with stat-
ic mixers overcomes these shortcomings by efficiently increasing the
⁎ Corresponding author. mass transfer rate between the phases without introducing moving
E-mail address: blaz.likozar@ki.si (B. Likozar). parts, intended for mixing (low investment/operating costs) [13,17,18].

http://dx.doi.org/10.1016/j.fuproc.2015.10.035
0378-3820/© 2015 Elsevier B.V. All rights reserved.
B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336 327

The background of the simulations for the continuous tubular reac-


tor with static mixers, used for biodiesel production, was initially pre-
sented by Santacesaria et al. [17,18]. While pioneering the field,
chemical kinetics were considered using glycerides as lumps, whereas
fluid mechanics, transport phenomena and intrinsic kinetics were not
treated simultaneously, in an overall reactor model, which is the
distinguishing innovation of this work.
In this study, chemical equilibrium, reaction kinetics and mass trans-
fer were modelled for the trans-esterification of oil to biodiesel in a con-
tinuous tubular reactor with static mixers at different temperatures,
volumetric flow rates, phase fractions, and catalyst contents. The
model was based on the distinguishability among various fatty acid-
bonded species and may thus be used for oil resources differing in com-
position. Furthermore, detailed sensitivity and error analyses were per-
formed, and the applicability of the developed model for the scale-up/
scale-down, economic optimization and intensification of the process
were investigated.

2. Material and methods

2.1. Materials

Methanol (99.8 wt.%, for liquid chromatography, containing


b0.03 wt.% water) was purchased from Merck (Darmstadt, Germany)
and canola oil from Tovarna olja Gea (Slovenska Bistrica, Slovenia, pur-
chased at a local food store). Hydrochloric acid (HCl, neutralizing acid)
used for the quenching of the chemical reactions was purchased from
Merck (Darmstadt, Germany). Potassium hydroxide (88 wt.%, reagent
grade), acetonitrile (99.9 wt.%, hypergrade for LC–MS), methanol
(99.8 wt.%, for liquid chromatography), n-hexane (98 wt.%, for liquid
chromatography) and isopropanol (99.9 wt.%, gradient grade for liquid
chromatography) were purchased from J.T. Baker (Deventer, Holland)
(catalyst) and Merck (Darmstadt, Germany) (solvents), respectively.

2.2. Continuous reactors

Three 18 mL, a 32 mL, and a 72 mL tubular reactor systems were con-


structed and are shown schematically in Fig. 1a and b. Polyacetal static
mixers (Koflo Corporation, Inc., Cary, IL, USA) were positioned inside
the reactors. The number of the elements comprising the mixers was
24. The inside diameters of the tubes and the outside diameters of the
mixers were 5 (3/16″), 6 (1/4″) and 10 (3/8″) mm, respectively. The
lengths of the polyacetal static mixers were 118 (4 5/8″), 152 (6″) and
241 (9 1/2″) mm, giving the surface areas of 0.00112, 0.00161 and
0.00447 m2 for the entire mixers, respectively. The reactors consisted
of a horizontally-positioned immersed Saint-Gobain (Courbevoie,
France) tube of 1.32 m length (5 mm internal diameter), 0.75 and
1.25 m length (6 mm internal diameter), and 0.32 and 1.30 m length
(10 mm internal diameter), which were connected to feed pumps.
Controlled-volume metering pumps (Thermo Fisher Scientific, Wal-
tham, MA, USA) were used to feed the oil and methanol/catalyst mix-
tures to the system, while combined hot-plate magnetic-stirrer
devices (IKA-Werke, GmbH & Co. KG, Staufen, Germany) were used to
heat and mix the oil and the mixtures at the speed of 800–1000 rpm.
A heat exchanger (Julabo Labortechnik, GmbH, Seelbach, Germany)
coupled with self-tuning electronic PID control software was used to
control the reaction temperature. Fig. 1. Overall schematic flow diagram of biodiesel production in experimental system (tu-
bular reactor with static mixers) (a), the tubular reactor with static mixers (b), and phases
2.3. Process conditions of reactions between oil, methanol, intermediates and products in the tubular reactor with
stoichiometric reactions for conversion of triglycerides to fatty acid methyl esters (c).

Experiments were carried out at 30, 40, 50, 60 and 65 °C in the 18 mL


(5 mm internal diameter) tubular reactor for up to 1 h, while the 0.2, 0.4, potassium hydroxide was investigated. Canola oil (different amounts)
0.6, 0.8, 1.0 and 1.2 wt.% (per oil weight) based concentrations of cata- was used in each run. Pressure was controlled at 0 kPa between the
lyst potassium hydroxide were investigated (see Table 1). In the other pump side and the output side of the reactor. All experiments and sam-
tubular reactors, experiments were carried out at 50 °C up to 1 h, ple analyses were carried out in the specified order to minimize the po-
while the 0.8 wt.% (per oil weight) based concentration of catalyst tential experimental difficulties due to e.g. differing catalyst contents
328 B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336

Table 1
Prediction of sample compositions and concentrations at different experimental conditions (reactor dimensions, temperature, residence time of liquid bulk/flow rate, phase ratio and cat-
alyst concentration).

Diam. Temp. Space time Res. time Flow rate Phase ratio Cat. conc. (wt.% TG conc. DG conc. MG conc. G conc.
(mm)/length (°C) (min) (min) (mL/min) (mol/mol) (per oil)) (kmol/m3oil) (kmol/m3oil) (kmol/m3oil) (kmol/m3oil)
(m)

5/1.32 50 9.0 10.1 2.0 6 0.8 0.248 0.088 0.100 0.678


5/1.32 50 4.5 4.8 4.0 6 0.8 0.355 0.159 0.122 0.461
5/1.32 50 3.0 3.2 6.0 6 0.8 0.428 0.204 0.127 0.330
5/1.32 50 2.2 2.4 8.0 6 0.8 0.484 0.230 0.124 0.245
5/1.32 50 1.8 1.9 10.0 6 0.8 0.530 0.246 0.115 0.188
5/1.32 50 1.5 1.6 12.0 6 0.8 0.568 0.253 0.106 0.149
5/1.32 30 2.2 2.4 8.0 6 0.8 0.911 0.135 0.009 0.016
5/1.32 40 2.2 2.4 8.0 6 0.8 0.748 0.207 0.049 0.071
5/1.32 60 2.2 2.4 8.0 6 0.8 0.280 0.139 0.145 0.533
5/1.32 70 2.2 2.4 8.0 6 0.8 0.197 0.051 0.095 0.761
5/1.32 50 2.2 2.4 8.0 3 0.8 0.505 0.238 0.121 0.219
5/1.32 50 2.2 2.4 8.0 4 0.8 0.494 0.234 0.122 0.232
5/1.32 50 2.2 2.4 8.0 5 0.8 0.488 0.232 0.123 0.240
5/1.32 50 2.2 2.4 8.0 7 0.8 0.482 0.229 0.124 0.249
5/1.32 50 2.2 2.4 8.0 8 0.8 0.480 0.229 0.124 0.251
5/1.32 50 2.2 2.4 8.0 6 0.2 0.748 0.229 0.045 0.040
5/1.32 50 2.2 2.4 8.0 6 0.4 0.626 0.256 0.088 0.101
5/1.32 50 2.2 2.4 8.0 6 0.6 0.544 0.249 0.112 0.173
5/1.32 50 2.2 2.4 8.0 6 1.0 0.440 0.210 0.127 0.311
5/1.32 50 2.2 2.4 8.0 6 1.2 0.405 0.190 0.127 0.370
6/0.75 50 4.5 5.2 4.0 6 0.8 0.354 0.159 0.122 0.464
6/0.75 50 3.0 3.4 6.0 6 0.8 0.427 0.203 0.127 0.332
6/0.75 50 2.3 2.5 8.0 6 0.8 0.483 0.230 0.124 0.247
6/0.75 50 1.8 2.0 10.0 6 0.8 0.528 0.245 0.116 0.190
6/1.25 50 9.0 10.1 3.6 6 0.8 0.252 0.091 0.101 0.670
10/0.32 50 4.5 6.5 4.0 6 0.8 0.359 0.162 0.122 0.454
10/0.32 50 3.0 4.0 6.0 6 0.8 0.432 0.206 0.127 0.323
10/0.32 50 2.3 2.9 8.0 6 0.8 0.489 0.232 0.123 0.239
10/0.32 50 1.8 2.3 10.0 6 0.8 0.534 0.247 0.115 0.184
10/1.30 50 9.0 10.2 8.0 6 0.8 0.250 0.090 0.100 0.673

because of thermal expansion. Several parallel runs were also per- preparation. The bulk of the resulting, predominantly homogeneous
formed (see Table SD.1, Supplementary data). Additional experiments mixture was transferred to a sample vial and diluted with 1 mL of
were conducted to verify the effect of methanol feed flow rate and the isopropanol/n-hexane (5:8 (v/v)). The bulk of the resulting mixture
content of the base catalyst, the mixture temperature and the phase was immediately analysed using high-performance liquid chromatog-
fraction. raphy (HPLC) according to the method used by Likozar and Levec [19,
20]. The oil in the feed stream before was also analysed by HPLC. The hy-
2.4. Process procedure drochloric acid solution was not added to the oil before the analysis by
HPLC as neutralization was not needed. The HPLC analysis of the same
The oil was pre-mixed and charged into the tubular reactor system samples after different settling times revealed that the neutralization
prior to each experimental run. Canola oil (18–72 mL) was charged method was effective, since no additional conversion of the samples
into the tubular reactor; the reactor was open and the feed pump was was found.
operating continuously. After 15–20 min of feed time, methanol and
the base catalyst were charged continuously into the tubular reactor 2.5. Analytical methods
with the feed pump at the flow rates of 0.44–2.44 mL/min. The heat ex-
changer was switched on before the oil pre-mixing, in order to achieve 2.5.1. Determination of concentrations of reactants, intermediates and
the reaction temperature (30, 40, 50, 60 and 65 °C). A thermocouple products
was used to monitor the reaction temperature. A stable reaction tem- An Agilent (Hewlett Packard) (Santa Clara, CA, USA) HPLC system
perature (± 0.02 °C) was achieved within 20 min for 30, 40, 50, 60 consisting of two HPLC pumps, a controller, a UV–Vis detector and an
and 65 °C after the starting of the heat exchange. The pressure was con- auto-sampler was used to analyse the contents of the output and
trolled at 0 kPa. The output products were collected in 8 mL vials. All ex- feed streams. Origin 8.1 SR3 software (OriginLab Corporation, North-
periments were conducted for up to 1 h. ampton, MA, USA) was utilized for analysis. The column used was
Additional experiments were performed to observe the effect of the 250 × 4.6 mm Synergi™ 4 μ Hydro-RP 80 Å, packed with 4 μm-
combined methanol/base catalyst and oil feed flow rate on the conver- particles with the pore size of 80 Å (Phenomenex, Torrance, CA,
sion for the base-catalysed transesterifications. These flow rates were 2, USA), connected in series with several filters. The mobile phases
4, 6, 8, 10 and 12 mL/min. were acetonitrile/methanol (4:1 (v/v)) (A) and isopropanol/n-
The output products, collected at specific times during the entire ex- hexane (5:8 (v/v)) (B) at the flow rate of 1.3 mL/min at 35 °C.
periment, were mixed with the maximal base content-equivalent Phase B was used to make the 6 mL/L sample solutions. One millilitre
amount of pro analysi hydrochloric acid (prepared in the 8 mL vials) of the solution was injected into the auto-sampler vials. Prior to each
and shaken by hand for about 3–5 s. This step served to stop any further analysis, the solutions were filtered through a polytetrafluoroethylene
reaction in the samples by neutralizing the base catalyst within the (PTFE) frit (Agilent, Santa Clara, CA, USA).
methanol/glycerol and oil/fatty acid methyl ester (FAME) phases, the The HPLC analysis was conducted according to the method present-
former containing most of the catalyst. The mixture was allowed to set- ed by Likozar and Levec [19,20]. Calibration curves were generated from
tle for at least 72 h and was again stirred by hand before further sample several standards; triglycerides (TG) (trilinolein, trilinolenin, triolein,
B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336 329

tripalmitin and tristearin), diglycerides (DG) (1,2-dilinolein, 1,3- diffuse from and to the oil phase (o). The differential mass balances
dilinolein, 1,2-dilinolenin, 1,3-dilinolenin, 1,2-diolein, 1,3-diolein, 1,2- for the methanol and oil phase for TG, DG, MG, G, M, and ME (OH− is
dipalmitin, 1,3-dipalmitin, 1,2-distearin and 1,3-distearin), monoglyc- the catalyst) can be written as follows, where cx,y, cx,y,i and c⁎x,y denote
erides (MG) (1-monolinolein, 2-monolinolein, 1-monolinolenin, 2- the bulk, interface and equilibrium concentration of component x in
monolinolenin, 1-monoolein, 2-monoolein, 1-monopalmitin, 2- phase y (mol/m3), Kc,x,y the overall mass transfer coefficient of compo-
monopalmitin, 1-monostearin and 2-monostearin) and FAME (methyl nent x, defined for phase y (m/s), kc,x,y the mass transfer coefficient of
gadoleate, linoleate, linolenate, myristate, oleate, palmitate and stea- component x in phase y (m/s), Dx,methanol/oil the distribution coefficient
rate). The injected standard concentrations were plotted against their of component x between the methanol and oil phase (/), jm,x the molar
peak area. Each standard was injected once at six different concentra- flux of component x (mol/(m2 s)), a the specific surface area (1/m), v
tions (the volumetric sample concentrations for TG, DG, MG and linear velocity, Dm the axial diffusivity of methanol drops in the oil
FAME were 1, 2, 3, 4, and 5 μL/mL (the injection volume was 20 μL)). phase (m2/s), Dx,o the axial diffusivity of component x in the oil phase
The calibration curves of the standard solutions showed good power (m2/s), x the position in the reactor, and t time (s).
law relationships. The retention times of the standards are shown in
∂cTG;m
2
∂ cTG;m ∂cTG;m  
Table SD.2 (Supplementary data). Fig. SD.1 (Supplementary data)
¼ Dm −v  K c;TG;m a cTG;m −cTG;m −k1 cOH− ;m cTG;m cM;m þ
shows typical chromatograms for a mixture of TG, DG, MG and FAME. ∂t ∂x 2 ∂x
þk2 cOH− ;m cDG;m cME;m
The fraction of a component in the mixture at time t, based on the in-
dividual fatty acid-balanced component concentration remaining/ ð2Þ
evolved in the reactor, was calculated from the following equation (Eq
∂cDG;m
2
∂ cDG;m ∂cDG;m  
(1) in a generalized manner illustrates the calculation procedure for ¼ Dm −v  K c;DG;m a cDG;m −cDG;m þ k1 cOH− ;m cTG;m cM;m −
the case when all bonded fatty acids would be of the same type, while ∂t ∂x2 ∂x
−k2 cOH− ;m cDG;m cME;m −k3 cOH− ;m cDG;m cM;m þ k4 cOH− ;m cMG;m cME;m
actually the fatty acid-based summation was made utilizing all compo-
ð3Þ
nents in Table SD.2, Supplementary data).
∂cMG;m
2
∂ cMG;m ∂cMG;m  
ncCOMPONENT ¼ Dm −v  K c;MG;m a cMG;m −cMG;m þ k3 cOH− ;m cDG;m cM;m −
X COMPONENT ¼ ð1Þ ∂t ∂x 2 ∂x
3cTG0 þ 2cDG0 þ cMG0 þ cAE0 −k4 cOH− ;m cMG;m cME;m −k5 cOH− ;m cMG;m cM;m þ k6 cOH− ;m cG;m cME;m
ð4Þ
XCOMPONENT was the fraction, and cTG0, cDG0, cMG0 and cME0 were the
initial concentrations of TG, DG, MG and FAME (in order to account for ∂cG;m
2
∂ cG;m ∂cG;m  
the presence of any DG, MG or FAME) in the reactor. cCOMPONENT was ¼ Dm −v ∓K c;G;m a cG;m −cG;m þ k5 cOH− ;m cMG;m cM;m −
∂t ∂x 2 ∂x
the concentration of a component left/evolved in the reactor after a spe- −k6 cOH− ;m cG;m cME;m
cific reaction time or space time (n is its bonded fatty acid number). ð5Þ

2.5.2. Determination of dispersed phase drop size ∂cM;m


2
∂ cM;m ∂cM;m  
¼ Dm −v ∓K c;M;m a cM;m −cM;m −k1 cOH− ;m cTG;m cM;m þ
The size of methanol drops was determined by digital imaging ∂t ∂x2 ∂x
þk2 cOH− ;m cDG;m cME;m −k3 cOH− ;m cDG;m cM;m þ k4 cOH− ;m cMG;m cME;m −k5 cOH− ;m cMG;m cM;m þ
(Nikon D3100, Tokyo, Japan) and consequent image processing (ImageJ,
þk6 cOH− ;m cG;m cME;m
Bethesda, MA, USA). Subsequently, the Sauter mean diameter was cal-
ð6Þ
culated at different times and positions along the reactor length.

∂cME;m
2
∂ cME;m ∂cME;m  
2.6. Biodiesel production process economics examination ¼ Dm −v ∓K c;ME;m a cME;m −cME;m þ k1 cOH− ;m cTG;m cM;m −
∂t ∂x2 ∂x
−k2 cOH− ;m cDG;m cME;m þ k3 cOH− ;m cDG;m cM;m −k4 cOH− ;m cMG;m cME;m þ k5 cOH− ;m cMG;m cM;m −
Biodiesel production process economics examination was per- −k6 cOH− ;m cG;m cME;m
formed, primarily describing cost analysis. Both primary and secondary ð7Þ
data was used to evaluate the costs; specifically, as primary data, equip-
ment dimensions (reactor and mixer lengths and diameters), operating ∂cTG;o
2
∂ cTG;o ∂cTG;o  
¼ DTG;o −v ∓K c;TG;o a cTG;o −cTG;o ð8Þ
conditions (temperature, phase fraction, catalyst concentration and ∂t ∂x 2 ∂x
total volumetric flow rates) and component concentrations (all glycer-
ides, esters, methanol, glycerol and catalyst) were used, while the sec- ∂cDG;o
2
∂ cDG;o ∂cDG;o  
¼ DDG;o −v ∓K c;DG;o a cDG;o −cDG;o ð9Þ
ondary data was related to the market prices of resources, products ∂t ∂x2 ∂x
and energy. The market prices of methanol, oil and KOH in September
2014 were 0.50, 0.76 and 1.1 $/kg, respectively. The corresponding ∂cMG;o
2
∂ cMG;o ∂cMG;o  
¼ DMG;o −v ∓K c;MG;o a cMG;o −cMG;o ð10Þ
price of energy was 2.03 $/J, while the pumping power was calculated ∂t ∂x 2 ∂x
according to pump specifications and efficiency.
∂cG;o
2
∂ cG;o ∂cG;o  
¼ DG;o −v  K c;G;o a cG;o −cG;o ð11Þ
3. Theory ∂t ∂x2 ∂x

The scheme of the continuous biodiesel production process is shown ∂cM;o ∂ cM;o
2
∂cM;o  
¼ DM;o −v  K c;M;o a cM;o −cM;o ð12Þ
in Fig. 1a and c. During the process, the reaction mixture passes through ∂t ∂x2 ∂x
the mass transfer-determining region (heterogeneous system
consisting of dispersed methanol and continuous oil phase) to the reac- ∂cME;o
2
∂ cME;o ∂cME;o  
¼ DME;o −v  K c;ME;o a cME;o −cME;o ð13Þ
tion kinetics-determining region (pseudo-homogeneous system ∂t ∂x2 ∂x
consisting of components in a single phase). In the first region, the reac-
tions, represented by the kinetic scheme (Fig. 1c; TG, DG, MG, G, M, ME, The differential mass balances of components simplify upon the
and OH− are tri-, di- and monoglyceride, glycerol, methanol, methyl transition to the pseudo-homogeneous regime (Eqs. (SD.1–6), Supple-
ester, and hydroxide ion (catalyst), respectively, while k1–k6 are the re- mentary data) (with the exclusion of mass transfer terms), where cx
action rate constants), occur only in the methanol phase (m) (where the and Dx denote the bulk concentration of component x and its axial
catalyst is present), while the reactants, intermediates, and products diffusivity.
330 B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336

It is well-known that at the beginning of transesterification reac- νo is the kinematic viscosity of the continuous phase, Reo its Reyn-
tions, the droplets of methanol are dispersed in vegetable oil phase; olds number, and d the unit-less diameter of reactor; the parameters
however, during the reactions' progress, emulsion may change to the of axial dispersion were acquired by regression analysis of the obtained
glycerol-containing droplets, dispersed in continuous FAME (or biodie- flow data according to Becker [25], and were C1 = 8.67, α = 0.543 and
sel) phase, and in addition, FAME and TG are to some extent miscible γ = 0.269. The measurement-determined and model-predicted Péclet
and form one single phase (described in the literature [21–24] and numbers and axial dispersion coefficients for the alcohol drops in oil
depicted in Fig. 1c). For model development, not many assumptions are presented in Table SD.3 (Supplementary data).
were taken into account with regard to phases and component distribu- Under a vast majority of experimental conditions, methanol phase
tion, but rather the actual state of the process was described, which may droplets could be observed at the outlet of tubular reactor, which
be in-depth elaborated as follows. Firstly, the last stage in Fig. 1c (glyc- means that the transition from the heterogeneous to the homogenous
erol phase separation) was not observed in the reactors at any time or operating regime did not occur. The longest residence time was approx-
reactor length, while secondly, the emulsion of methanol in oil (the sec- imately 18 min, whereas the transition to the homogenous system in
ond step in Fig. 1c), as well as pseudo-homogeneous regime (the third batch-wise operation was reported to occur mostly in less than 3 min
step in Fig. 1c) were individually described by Eqs. (2–13) and [19]. Nonetheless, it has to be noted that the bulk of the reaction mixture
Eqs. (SD.1–6), Supplementary data) for heterogeneous and homoge- in batch operation was predominantly subjected to a much higher shear
neous system, respectively. Also, no assumptions were made regarding stress (e.g. when using 200–600 rpm), whereas at very low mixing in-
the transition between these two regimes and the droplet size in the tensities (100 rpm), the extent of reactions remained under 5% even
former one, as the system was monitored and visually recorded at all as long as 75 min [19]. Analogously, in the current continuous operation,
times and positions, while the actual parameters (Section 2.5.2) were static mixers allowed for a good individual homogenisation of both con-
input to the model. Regarding the distribution of the components be- tinuous and dispersed phase to decrease the constraints, superimposed
tween methanol and oil phase, the latter was calculated for each com- by transport phenomena (and hence achieve a reasonably high conver-
ponent (Table SD.2, Supplementary data) individually, based on the sion), but not for a high-enough shear stress to adequately increase the
existing distribution coefficients [19,20]. The mentioned glycerol interface area to usher in ME emulsification, and the ultimate homoge-
phase separated from biodiesel phase only upon collecting the product neous system transition.
in reactor outlet and leaving it to settle, thus obtaining, for example, bio- Reactions should therefore take place in the heterogeneous, mass
diesel to be subjected to final characterization. transfer-determining region. Interestingly, simulations' results showed
that mass transfer limitations were negligible, which is often the case
4. Results and discussion with two-phase droplet systems. The circulation of flow inside droplets,
as well as the mixing effect these droplets exert on bulk flow, along with
4.1. Modelling mass transfer and reaction kinetics of basic chemical reac- enhanced mixing due to the presence of static mixers, produces a well-
tions considering fatty acid composition for biodiesel production with ho- mixed system where the interfacial film thickness is reduced and bulk
mogeneously catalysed transesterification concentrations are equal to the interface ones. After the adjustment of
mass balance equations (Eqs. (2–13)) by excluding mass transfer limita-
The partial differential mass balance equations for glycerol and tions, which were proven negligible, no fitting procedures were in-
methanol, as well as for each individual TG, DG, MG and ME inside the volved in the calculations.
oil and alcohol phase, were discretized in accordance with the finite dif- Continuous operation therefore somewhat deviates from (resi-
ference methodology and solved inside the computational domain of dence) time-analogous batch processing, which was investigated in
space and time. Grid independence was achieved with a 100 × 1400- our previous works [19,20]. Under batch operation, the conversion of
element length and time discretization. Calculations were performed oil was negligible during the initial heterogeneous, mass transfer-
in MATLAB (MathWorks, Natick, MA, USA) and the corresponding sim- influenced region, and a considerable conversion was noticed only
ulations were executed under the same operating conditions as the after the transition to the homogeneous, reaction kinetics-determined
(transient/stationary) experiments, which are presented in Table SD.1 phase. This emphasizes the significance of using effective static mixers,
(Supplementary data). such as the ones in this study (called Kenics, KM or helical static mixers),
The kinetic parameters were obtained from the work by Likozar and as an efficient process intensification strategy. This is contradictory to
Levec [19,20]. In the mentioned work, the parameters were acquired by the observation made by Qiu et al. [2], who concluded that the mixing
the regression analysis of the experimental data obtained from batch process by static mixers relies on slow, unforced molecular inter-
runs. They were determined through the modelling, based on the mech- diffusion (the general flow pattern plays a decisive role).
anism and reaction scheme of individual TG, DG, MG, and ME, contain- Fig. 2a–d present the measured (symbols) and simulated (lines)
ing different combinations of gadoleic, linoleic, linolenic, oleic, palmitic fatty acid chain-based fractions of TG, DG, MG and ME (Eq. (1)) at the
and stearic acids [19,20]. Their boiling points, molar volumes at boiling reactor outlet versus time for different flow rates. The operating condi-
points, and molar volumes and viscosities versus temperature for the tions for the 1.317 m-long tubular reactor with the inner diameter of
considered components can be found in Fig. SD.2 (Supplementary 5 mm were 50 °C, the methanol/oil molar ratio of 6 and using 0.8 wt.%
data), while the reaction rate constants (and implicitly, pre- KOH catalyst. Experimental points were obtained at specific sampling
exponential factors and activation energies) for the fractional conver- times, while simulations show the predicted conversions for the com-
sion of different TG, DG and MD versus the reciprocal temperature for plete experimental time scale. At lower flow rates, conversions were
the reactions in single/two phases are presented in Fig. SD.3 (Supple- higher, and hence, the fatty acid chain-based fractions of TG (Fig. 2a),
mentary data). The pre-exponential factors and activation energies DG (Fig. 2b) and MG (Fig. 2c) were lower, while the ones of ME
were also related with the structure of reactants (oil/alcohol), interme- (Fig. 2d) were higher due to longer residence times and (obviously)
diates and products, acknowledging the number of carbons, double non-equilibrium conditions. Soon after the passing of 1 residence
bonds and alkyl branches by linear and mixed response surface meth- time, steady state was practically achieved, which is represented by
odology [19,20]. the plateau at higher process times. At the flow rate of 2 mL/min, the
The axial dispersion of the methanol phase (droplets) was defined ME fraction was almost 70% (Fig. 2d). Fig. 2e and f show the individual
with the following equation. concentrations of TG and DG at the above-mentioned operating condi-
tions and at the flow rate of 8 mL/min. Conversely to Fig. 2, reactors
were filled with oil prior to the start of each experiment, but the mea-
γ
Dm ¼ ν o C 1 Reαo d ð14Þ sured values were recalculated for a better representation, as if the
B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336 331

Fig. 2. TG (a), DG (b), MG (c) and ME (d) fractions, and individual TG (e) and DG (f) concentrations versus total elapsed time for the two-phase reaction system at the temperature of 50 °C,
phase fraction 1:6 (mol/mol) (oil to methanol), catalyst concentration 0.8 wt.% (per oil weight), and different total volumetric flow rates (1.9, 3.7, 5.6, 7.5, 9.4 and 11.4 × 10−3 m/s). Reactor
length and volume were 1.32 m and 18 mL, while the diameter and length of packing static mixer were 5 and 118 mm, respectively. For all figures, solid, dotted and dashed lines represent
model simulations, and symbols represent the experiments performed at different measurement conditions.

reactors would have been filled with an inert oil-like fluid. For the sim- Fig. 3 shows the results, obtained at different space times (corre-
ulations, the initial conditions were therefore set to zero concentration sponding to the same experimental conditions as in Fig. 2), which
inside the reactor prior to commencement. The steady-state solution were calculated by dividing the void reactor volume with the total vol-
is the same for both cases, whereas a very good fit was obtained, umetric flow rate. Fig. 3a and b therefore show how the fatty acid chain-
refrained from any parameter regression for the measured data. based fractions of TG, DG, MG and ME change with varying flow rate. TG
The effect of axial dispersion can be observed as the initial slower are solely depleted during the reactions, ME are produced (Fig. 3a),
concentration rise, just before t = τ, and as the deceleration before while DG and MG are both produced and subsequently depleted
reaching steady state (also seen in component fraction figures), this (Fig. 3b), according to the chemical kinetics of the reactions (Fig. 1c).
being the most profound at lower flow rates. High flow rates produce Fig. 3c–f present the individual concentrations of the components (reac-
a steep breakthrough, which is typical for evolved plug flow. The full tants, intermediates and products). The concentrations of TG decrease
concentration profiles of all the components versus length and time with time as a rule (Fig. 2c), while the concentrations of DG initially
are presented in Fig. SD.4 (Supplementary data). rise, as they are produced (some are already present in oil and only
332 B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336

Fig. 3. TG (a), DG (b), MG (b) and ME (a) fractions, and individual TG (c), DG (d), MG (e) and ME (f) concentrations versus static mixer-packed reactor space time for the two-phase re-
action system at the temperature of 50 °C, phase fraction 1:6 (mol/mol) (oil to methanol), catalyst concentration 0.8 wt.% (per oil weight), and different total volumetric flow rates (1.9, 3.7,
5.6, 7.5, 9.4 and 11.4 × 10−3 m/s). Reactor length and volume were 1.32 m and 18 mL, while the diameter and length of packing static mixer were 5 and 118 mm, respectively.

exhibit a decrease due to the prevailing forward reactions), only to rate of olein-containing species, observable as less sharp peaks for inter-
lower successively as they are depleted (Fig. 3d). MG concentrations mediate components. By following the evolution and depletion of indi-
(Fig. 3e) show a similar trend, while the production of ME (biodiesel) vidual species, one can identify reactions and components, which limit
increases monotonously with the residence of the reaction mixture. the rate of biodiesel production, and in this sense, we can choose the op-
In Fig. 3e and f, a rapid conversion of Ln, P and S into MLn, MP and MS timal initial oil constitution for efficient biodiesel production. For in-
can be noticed, since the concentrations of these monoglycerides stance, in Fig. 3c one can observe how slowly OOO, OOP or LLS are
are very low throughout the experiments. Nonetheless, the comparative depleted. It is unfortunate that some of these components are rather
evolution of MO as per the most abundant ML is rather steadier when abundant in oil (some of these components have a high initial concen-
applying the straightforward time–space time analogy with batch tration). On the other hand, the reactions involving LLL and OLL are
runs [19]. The latter is a consequence of a lower overall alcohol availabil- fast, as a rapid drop in concentration can be observed. It is advantageous
ity as a reactant, as conversely to batch runs, the space time process to produce biodiesel from oils with a high content of these TG. Likewise,
window is mainly bi-phasic in nature, decreasing especially the in Fig. 3d, one can see that SS is unwilling to convert efficiently, which
B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336 333

would render one reluctant to use oil with a high SSS or SS content. The axial dispersion. The effect of temperature is significant (Fig. 4a); elevat-
produced biodiesel from the applied commercial canola oil comprised ed temperatures are highly beneficial for biodiesel production, as about
55.1% ML, 28.5% MO, 11.9% MLn, 3.6% MP, 0.7% MS and 0.2% MG. 70% conversion is achieved at 65 °C, whereas the conversion is less than
5% at 30 °C, at comparable conditions. Lower flow rates provide a larger
4.2. Predicting mass transfer and reaction kinetics upon differing main fac- residence time for achieving higher conversions (Fig. 4b); the maxi-
tors affecting the yield of biodiesel: alcohol quantity, reaction time, reaction mum (equilibrium) conversion, which can be achieved by using utterly
temperature and catalyst concentration slow flow rates (or a longer reactor tube), was identified at approxi-
mately 87%, which is in accordance with the observations, made in the
Fig. 4a–d present the fatty acid chain-based fractions at different literature [7]. Lower flow rates are, on the other hand, unfavourable
temperatures, volumetric flow rates, phase fractions and catalyst con- for achieving a high productivity. The fatty acid chain-based ME frac-
centrations, in the same order. Steady state results are shown, as well tions around τ follow an oscillating curve — this is due to
as the transient data at 0.9, 1.0 and 1.1 space time (τ). The conversions experimentally-determined droplet diameters (Section 2.5.2.) (and
around τ are lower than steady state analogues due to the influence of their corresponding errors) being (in)directly fed into the model (Dm

Fig. 4. Effect of reaction temperature (a,e,f), volumetric flow rate (b), phase fraction (c) and base-catalyst concentration (d) (non-linear model simulations plotted for all measurement
conditions at steady state (a–f) and during the transient conditions at 0.9τ, τ and 1.1τ (a–d)). 50 °C, 1:6 (mol/mol) (oil to methanol), 0.8 wt.% (per oil weight), and 8 mL/min were
used unless specified otherwise, while reactor length and volume were 1.32 m and 18 mL, respectively.
334 B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336

and a in Eqs. (2–13)), which causes the variations of component con- concentration of the component with the highest concentration (OLL)
centrations, which can be observed at the exit of the tubular reactor was appropriately adjusted, so that the overall concentration of oil com-
around τ, yet, the latter stabilizes upon approaching steady state. ponents remained the same. In the case of OLL, the concentration of the
Fig. 4c depicts the effect of methanol/oil phase ratio, conversion ris- second most abundant oil component (OOL) was adjusted accordingly.
ing rapidly with an increasing methanol/oil ratio up to the value of 4, ME conversions (fatty acid chain-based ME fractions) are illustrated, the
after which it remains approximately the same. For this set of experi- designation “Original” representing the source ME conversion in the ex-
mental conditions, the optimal methanol/oil phase ratio is therefore 4, periment at the temperature of 50 °C, phase ratio of 1:6 (mol/mol) (oil
and adding more methanol to the system does not increase the final to methanol), catalyst concentration of 0.8 wt.% (per oil weight), and
ME fraction. While in batch mode, one can observe a further influence the flow rate of 8 mL/min.
of the phase ratio up to the value of 7, when it gradually ceases to affect The bar chart identifies the components which most profoundly in-
the rate and yield [19]; the rationale for the plateau-like behaviour lies fluence ME conversion. For instance, a 20% higher initial concentration
in the interplay of reaction kinetics and thermodynamics. In batch oper- of LLL in the oil feedstock induces a 3% raise in ME conversion. After
ation, the differences upon the increase of the phase ratio typically be- LLL, which has the biggest positive effect, OLL also positively influences
come more profound with the time lapse (i.e. closing equilibrium), ME conversion. LLL and OLL are the components which react the most
and this can to some extent be potentially translated to the residence readily, as can also be seen in Figs. 3c (and 4e), where the two compo-
time in Fig. 4c as well. Higher catalyst concentrations result in higher nents exhibit a rapid drop. Fig. 3e depicts a correspondingly high L
conversions to fatty acid methyl esters (Fig. 4d). At the catalyst concen- and O production, while Fig. 3f the ultimate end-product, ML and MO
tration of 1.2 wt.%oil, the conversion was the highest; this observed evolution, the latter being the main components in the produced
trend would continue, only to gradually decrease in impact as can be biodiesel.
seen in Fig. 4d and in the batch data [19]. On the other hand, a 20% initial raise in the concentrations of OOL
The last two figures (Fig. 4e and f) present the concentrations of TG and OOO causes a negative effect on conversion. A 20% higher initial
and DG at the outlet of the reactor at various temperatures. Once again, OOL concentration induces a 2% lowering of ME conversion. OOL and
we can notice how OOO, OOP, LLS and SS do not readily convert even at OOO are both present at high levels in canola oil, used for experiments,
elevated temperatures; virtually unchanged concentrations at all tem- as can be seen in Fig. 3c. After 10 min space time, the concentrations of
peratures indicate a very low conversion. Oil, which has a high fraction these 2 components remain high and are unconverted. Otherwise, the
of these components, is thus less suitable for high-yield biodiesel variations of the initial amounts of most components do not cause any
production. significant differences. Fig. 5d shows an opposite effect, specifically,
when the initial concentrations are 20% lower. A lower initial OOL and
4.3. Scale-up and scale-down of biodiesel production process OOO amount therefore benefits ME conversion, correspondingly to
Fig. 5c.
For the assessment of the characteristics of process scale-up, two
other reactors with a larger inner diameter were experimented on. 4.5. Optimization and process intensification of biodiesel production
The first had the inner diameter of 6 mm, and the length of 0.75 m, process
while the other had the inner diameter of 10 mm and the length of
0.32 m, respectively. The void volumes of both reactors were the same The total cost of biodiesel production (C) was calculated for the case,
as in the case of the narrowest-tube reactor (5 mm) used for the when the phase ratio of 1:6 (mol/mol) (oil to methanol), the catalyst
above-mentioned experiments (18 mL). The static mixers used were concentration of 0.8 wt.% (per oil weight), and the total (oil and metha-
appropriately larger, with their diameter corresponding to the inner di- nol) volumetric flow rate of 8 mL/min were applied. Reactor length and
ameter of the reactors. An efficient scale-up from lab- to pilot plant volume were 1.32 m and 18 mL, respectively. Production cost was calcu-
level, or even commercial scale, is essential for achieving a higher bio- lated for the temperatures in the range 30–70 °C, according to the fol-
diesel capacity. A larger inner reactor tube diameter influences the lowing expression.
axial dispersion of bulk phase components and methanol phase itself 
(Eq. (14)), the diameter having a positive effect on the increase of dis- C ¼ Q o ρo C p;o þ Q m ρm C p;m ΔTC energy þ PC energy þ
ð15Þ
persion. The influence is exerted directly through d; nonetheless, also þQ o ρo C o þ Q m ρm C m þ Q o ρo wKOH C KOH
through the flow pattern (Re number). A higher axial dispersion can
lower the conversion of reactive components, while the concentration Qi is the volumetric flow rate of either oil (o) or methanol (m), ρi
variations in the radial direction also come in effect at larger (industrial their density, cp heat capacity, ΔT the difference between operational
reactor) diameters, which is also undesirable. and ambient temperature, Cenergy the cost of energy, P pumping
When dealing with inner diameter increase, the effect of gravity or power, Co the price of oil, Cm the price of methanol, wKOH the weight
mass transfer limitations due to larger droplets having a smaller specific fraction of KOH catalyst with respect to oil weight, and CKOH the price
surface area may lead to the deviations from ideal (model-predicted) of KOH catalyst.
behaviour, leading to a nonhomogeneous concentration distribution. Results are presented in Fig. 6. The cost of pumping is the lowest by
The model in Fig. 5a and b predicts practically identical conversions many orders of magnitude as compared to the other items. The cost of
for all 3 reactors. This implies that an efficient scale-up is possible, ex- heating increases with temperature, but is still 2–3 orders of magnitude
perimental data confirming these findings, which means that any lower than the one of oil and catalyst, which are the most pricy, while
scale-down to smaller dimensions (e.g. micro-reactors) is unnecessary the cost of methanol is still somewhat lower. Since the majority of the
at these conditions, since a large-enough surface area is achieved even expenses for biodiesel production originate from feedstock price,
in a 10 mm-inner diameter tubular reactor. Nonetheless, is has to be which de facto dictates the costs of this fuel [19], the operation at 65
pointed out that the noted independence of tube diameter is not antic- °C is preferable, since it does not add significantly to the total cost, but
ipated to apply indefinitely. at the same time profoundly increases the reaction rates, and with it,
the overall achievable process productivity. In addition, costs were
4.4. Sensitivity and error analysis of biodiesel production process also determined per a litre of the produced biodiesel for the comparison
purposes with other works and process configurations, and were
Fig. 5c and d presents the sensitivity analysis of biodiesel production, established as 1.219, 1.211, 1.201, 1.191 and 1.183 $/L for the individual
which considers a +20% (Fig. 5c) and a −20% (Fig. 5d) variation of each cases, presented in Fig. 6, and the temperatures of 30, 40, 50, 60 and 70 °C,
oil component, present in feedstock. Upon each differing, the correspondingly.
B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336 335

Fig. 5. TG (a), DG (b), MG (b) and ME (a) fractions, versus static mixer-packed reactor space time, and the ME fraction upon increasing (c) or decreasing (d) the individual initial TG or DG
concentration by 20% at for the reactions at the temperature of 50 °C, phase fraction 1:6 (mol/mol) (oil to methanol), catalyst concentration 0.8 wt.% (per oil weight), and different total
volumetric flow rates (1.9, 3.7, 5.6, 7.5, 9.4 and 11.4 × 10−3 m/s in Reactor 1, and appropriately lower for Reactors 2 and 3). Reactor length and volume were 1.32 m (Reactor 1), 0.75 m
(Reactor 2) and 0.32 m (Reactor 3), and 18 mL (Reactors 1–3), while the diameter and length of packing static mixer were 5 and 118 mm (for Reactor 1), 6 and 152 mm (for Reactor 2), and
10 and 241 mm (for Reactor 3), respectively.

The quality assessment of the produced biodiesel was also per-


formed according to EN 14214 or ASTM D6751. Some of the most vital
parameters, listed in these norms, selected according to technical crite-
rion, include ester content, the density at 15 °C, the viscosity at 40 °C and
others. The ester content determined was 93.8%, which is a bit below
the standard lower limit of 96.5%; nonetheless, the examined sample
was the one, obtained at central experimental conditions, not the opti-
mal ones, which would allow for surpassing the mentioned limit. The
density at 15 °C of 883.5 kg/m3 and the kinematic viscosity at 40 °C of
3.773 mm2/s were well within the boundaries of 860–900 kg/m3 and
3.5–5.0 mm2/s, respectively. Acid value was lower than 0.05 mgKOH/g,
compared to the upper limit of 0.05 mgKOH/g, while the flash point
was higher than 155 °C, considering the lower limit of 101 °C.

5. Conclusions

Continuous biodiesel production was performed in the two-phase


regime with mass transfer being a non-determining step. Modelling re-
sults confirmed that the two-phase system was well mixed and that the
Fig. 6. Biodiesel production costs versus temperature at the phase fraction of 1:6 (mol/mol) interfacial concentrations of components were identical to their bulk
(oil to methanol), catalyst concentration 0.8 wt.% (per oil weight), and total (oil and metha-
nol) volumetric flow rate 8 mL/min (the superficial velocity in the empty reactor of
counterparts. The concentration profiles of each individual TG, DG, MG
7.5 × 10−3 m/s). Reactor length and volume were 1.32 m and 18 mL, while the diameter and ME were established based on their fatty acid composition, the lat-
and length of packing static mixer were 5 and 118 mm, respectively. ter not being undertaken beforehand.
336 B. Likozar et al. / Fuel Processing Technology 142 (2016) 326–336

Process modelling allowed for the identification of the components [6] F. Ataya, M.A. Dube, M. Ternan, Transesterification of canola oil to fatty acid methyl
ester (FAME) in a continuous flow liquid–liquid packed bed reactor, Energy Fuel 22
within the oil feedstock, which are readily converted, and the ones, (2008) 3551–3556.
which are not that prone to react. It was shown that it is advantageous [7] D.Y.C. Leung, X. Wu, M.K.H. Leung, A review on biodiesel production using catalyzed
to use oil with a high OOL, OLL or LLL content, while it is also desired to transesterification, Appl. Energy 87 (2010) 1083–1095.
[8] A.P. Harvey, M.R. Mackley, T. Seliger, Process intensification of biodiesel production
have a low OOO, LLS and OOP fraction. Higher conversions to fatty acid using a continuous oscillatory flow reactor, J. Chem. Technol. Biotechnol. 78 (2003)
methyl esters were predicted, simulating the initial LLL and OLL content, 338–341.
increased by 20%. Conversely, a 20% higher initial oil fraction of OOL or [9] J.L. Massingill, P.N. Patel, M. Guntupalli, C. Garret, C. Ji, High efficiency nondispersive
reactor for two-phase reactions, Org. Process. Res. Dev. 12 (2008) 771–777.
OOO exhibited an adverse effect. For the case of the applied canola oil,
[10] P.C. Narvaez, F.J. Sanchez, R.D. Godoy-Silva, Continuous methanolysis of palm oil
the biodiesel produced comprised mainly ML, MO, MLn and MP. In- using a liquid–liquid film reactor, J. Am. Oil Chem. Soc. 86 (2009) 343–352.
creasing the temperature had a drastic positive effect on the yield, [11] M.N. Kashid, L. Kiwi-Minsker, Microstructured reactors for multiphase reactions:
state of the art, Ind. Eng. Chem. Res. 48 (2009) 6465–6485.
while the optimal methanol to oil ratio was identified to be 4. The tubu-
[12] N. Sdrula, A study using classical or membrane separation in the biodiesel process,
lar reactor with the largest diameter examined (10 mm) provided virtu- Desalination 250 (2010) 1070–1072.
ally identical results, which means that efficient scale-up is possible. [13] B. Klofutar, J. Golob, B. Likozar, C. Klofutar, E. Žagar, I. Poljanšek, The
Ultimately, the production cost assessment showed that it is beneficial transesterification of rapeseed and waste sunflower oils: mass-transfer and kinetics
in a laboratory batch reactor and in an industrial-scale reactor/separator setup,
to operate at higher temperatures, since the majority of the expendi- Bioresour. Technol. 101 (2010) 3333–3344.
tures come due to oil and catalyst market price. [14] G.Q. Guan, M. Teshima, C. Sato, S.M. Son, M.F. Irfan, K. Kusakabe, N. Ikeda, T.J. Lin,
Two-phase flow behavior in microtube reactors during biodiesel production from
waste cooking oil, AICHE J. 65 (2010) 1383–1390.
Acknowledgements [15] E.E. Kalu, K.S. Chen, T. Gedris, Continuous-flow biodiesel production using slit-
channel reactors, Bioresour. Technol. 102 (2011) 4456–4461.
The provision of financial support for the conduct of the research and [16] X.H. Yu, Z.Z. Wen, Y. Lin, S.T. Tu, Z.D. Wang, J.Y. Yan, Intensification of biodiesel syn-
thesis using metal foam reactors, Fuel 89 (2010) 3450–3456.
preparation of the article by Slovenian Research Agency (ARRS) (Pro- [17] E. Santacesaria, M. Di Serio, R. Tesser, M. Tortorelli, R. Turco, V. Russo, A simple de-
gramme P2-0152) is gratefully acknowledged. The authors would also vice to test biodiesel process intensification, Chem. Eng. Process. 50 (2011)
like to thank Martin Petric for providing experimental help during 1085–1094.
[18] E. Santacesaria, R. Turco, M. Tortorelli, V. Russo, M. Di Serio, R. Tesser, Biodiesel pro-
research. cess intensification by using static mixers tubular reactors, Ind. Eng. Chem. Res. 51
(2012) 8777–8787.
Appendix A. Supplementary data [19] B. Likozar, J. Levec, Effect of process conditions on equilibrium, reaction kinetics and
mass transfer for triglyceride transesterification to biodiesel: experimental and
modeling based on fatty acid composition, Fuel Process. Technol. 123 (2014)
Supplementary data to this article can be found online at http://dx. 108–120.
doi.org/10.1016/j.fuproc.2015.10.035. [20] B. Likozar, J. Levec, Transesterification of canola, palm, peanut, soybean and sun-
flower oil with methanol, ethanol, isopropanol, butanol and tert-butanol to biodie-
sel: modelling of chemical equilibrium, reaction kinetics and mass transfer based on
References fatty acid composition, Appl. Energy 122 (2014) 30–41.
[21] M.A. Dubé, A.Y. Tremblay, J. Liu, Biodiesel production using a membrane reactor,
[1] M.K. Lam, M.T. Lee, A.R. Mohamed, Homogeneous, heterogeneous and enzymatic Bioresour. Technol. 98 (2007) 639–647.
catalysis for transesterification of high free fatty acid oil (waste cooking oil) to bio- [22] P. Cao, M.A. Dubé, A.Y. Tremblay, Methanol recycling in the production of biodiesel
diesel: a review, Biotechnol. Adv. 28 (2010) 500–518. in a membrane reactor, Fuel 87 (2008) 825–833.
[2] Z.Y. Qiu, L.N. Zhao, L. Weather, Process intensification technologies in continuous [23] M.C.S. Gomes, N.C. Pereira, S.T.D. de Barros, Separation of biodiesel and glycerol
biodiesel production, Chem. Eng. Process. 49 (2010) 323–330. using ceramic membranes, J. Membr. Sci. 352 (2010) 271–276.
[3] L. Xiang, L. Wang, J. Cheng, G.H. Que, Continuous production of biodiesel from soy- [24] I. Reyesa, G. Ciudad, M. Misra, A. Mohanty, D. Jeison, R. Navia, Novel sequential batch
bean oil and methanol in an enforced mass transfer reactor, Energy Sources, Part A membrane reactor to increase fatty acid methyl esters quality at low methanol to oil
33 (2011) 859–868. molar ratio, Chem. Eng. J. 197 (2012) 459–467.
[4] C. Da Silva, F. de Castilhos, J.V. Oliveira, L.C. Filho, Continuous production of soybean [25] O. Becker, Axial mixing and scaleup in static packed liquid extraction columns,
biodiesel with compressed ethanol in a microtube reactor, Fuel Process. Technol. 91 Chem. Eng. Technol. 26 (2003) 35–41.
(2010) 1274–1281.
[5] M.A. Dube, A.Y. Tremblay, J. Liu, Biodiesel production using a membrane reactor,
Bioresour. Technol. 98 (2007) 639–647.

You might also like