Date post: | 25-Nov-2023 |
Category: |
Documents |
Upload: | teknologimalaysia |
View: | 0 times |
Download: | 0 times |
Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx
G Model
JTICE-654; No. of Pages 8
Immobilized lipase-catalyzed transesterification of Jatropha curcas oil:Optimization and modeling
Alireza Zarei a, Nor Aishah Saidina Amin a,*, Amin Talebian-Kiakalaieh a, Nor Azimah Mohd Zain b
a Chemical Reaction Engineering Group (CREG), Faculty of Chemical Engineering, Universiti Teknologi Malaysia, UTM, 81310 Skudai, Johor, Malaysiab Department of Industrial Biotechnology, Faculty of Biosciences and Bioengineering, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia
A R T I C L E I N F O
Article history:
Received 29 January 2013
Received in revised form 8 May 2013
Accepted 16 May 2013
Available online xxx
Keywords:
Jatropha oil
Biodiesel
Transesterification
Immobilized lipase
Optimization
A B S T R A C T
This study investigates the production of biodiesel from non-edible crude jatropha oil in the presence of
an immobilized lipase catalyst. A modified method was used to immobilize Rhizopus oryzae lipase in the
polyvinyl alcohol–alginate matrix. Response surface methodology (RSM) and artificial neural network
(ANN) were employed to evaluate the relationship between process variables and biodiesel yield and
predict the optimal reaction conditions. The determination coefficient values for RSM and ANN were 0.98
and 0.97, respectively, indicating that both models can accurately predict the experimental results. The
experimental results revealed that the highest biodiesel yield was 87.10% at 40 8C reaction temperature,
5:1 methanol/oil molar ratio, 70 wt% water content, and 17 h reaction time. The immobilized lipase-
catalyst beads exhibited good activity for biodiesel production, as indicated by the major properties that
complied with the American Society for Testing and Materials (ASTM) international D6751 standard.
� 2013 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Contents lists available at SciVerse ScienceDirect
Journal of the Taiwan Institute of Chemical Engineers
jou r nal h o mep age: w ww.els evier . co m/lo c ate / j t i c e
1. Introduction
Fatty acid methyl ester (FAME) compounds, also known asbiodiesel because of their properties similar to diesel fuel, havepotential to be the main fuel source in the near future [1–3].Biodiesel is biodegradable, technically feasible, and environmen-tally friendly renewable fuel [3–5].
Several sources are available as feedstock for biodieselproduction, such as animal fats and vegetable oils [6]. Vegetableoils are more widely used than other sources, and are divided intotwo categories: (1) edible oils, such as soybean [7], palm [8],sunflower [9], and coconut [10]; and (2) inedible oils, such as wastecooking oil (WCO) [11], algae [12], and some bean oils, such asJatropha curcas (Jatropha) [13], Pongamina pinnata (Karanja) [14],Hevea brasiliensis (Rubber) [15], and Calophyllum inophyllum
(Polanga) [16].Biodiesel production using edible oils has been controversial
because it threatens human food sources worldwide [3,17,18],thereby increasing the prices of edible oils and consequently, theprice of biodiesel [19]. About 350 oil-bearing crops have beenrecognized as potential biodiesel sources [20]. Among which,Jatropha, is one of the best candidates that possess resistance
* Corresponding author. Tel.: +60 7 553 5579; fax: +60 7 558 8166.
E-mail addresses: [email protected] (A. Zarei), [email protected],
[email protected] (N.A.S. Amin).
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
1876-1070/$ – see front matter � 2013 Taiwan Institute of Chemical Engineers. Publis
http://dx.doi.org/10.1016/j.jtice.2013.05.015
against drought and can restore environmental wastelands[18,21].
The most common method for biodiesel production is vegetableoil alcoholysis or transesterification reaction [5,21]. In this method,vegetable oil reacts with alcohol in the presence of a catalyst or insupercritical conditions that result in the production of glyceroland biodiesel [4,5].
Acid [4], base [22], and biocatalyst [1] have been used in thecatalytic transesterification process. Enzymes are considered asbiocatalysts and can improve the chemical reactions within livingcells by lowering the activation energy [23]. Compared withchemical catalysts in the transesterification reaction, biocatalystshave the following advantages: (1) environmentally friendly, (2)exhibit higher reaction rates under ambient conditions, (3) havelittle or no chemical decomposition during the process, (4) requireless alcohol, and (5) have purer final product.
Immobilized forms or free enzymes may be used in industries.Immobilization enhances the stability of enzymes against variousdenaturing forms and also immobilized enzyme separation at theend of the reaction will be direct [24]. In the 1960s, enzymeentrapment techniques were implemented using synthetic poly-mer gels, such as polyacrylamide gel or polyvinyl alcohol (PVA), ornatural polymer derivatives such as silicon elastomers for the sol–gel process, starch, or nitrocellulose [25]. Enzyme-adsorptivecross-linking on membranes and films or beads for the formationof enzyme envelopes has also been developed. Immobilizedbiocatalysts gained stable interests during the past decade,
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
hed by Elsevier B.V. All rights reserved.
Table 2Experimental range and levels of independent variables in transesterification
reaction.
Independent variable Range and level
�a �1 0 +1 +a
A: Time (h) 1 5 9 13 17
B: Temperature (8C) 30 35 40 45 50
C: Methanol/oil molar ratio 3:1 4:1 5:1 6:1 7:1
E: Water content (wt%) 10 40 70 100 130
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx2
G Model
JTICE-654; No. of Pages 8
significantly increasing the number of current scientific reports inbiotechnology related journals. Tamalampudi et al. [26], havecompared R. oryzae immobilized lipase with one of the mostcommon biocatalyst for biodiesel production, Novozym 435. Theyshowed that R. oryzae lipase is more efficient than Novozym 435,regardless of alcohol type.
Process optimization is important and helpful for the industri-alization and development of biodiesel production. RSM-utilizingstatistical methods can provide a research strategy for studyingparameter interaction; thus, biodiesel production in differentconditions, with minor estimation error, can be modeled with RSM[27]. By utilizing this technique, existing data can be used to avoidtrial runs by clarifying the reaction conditions [28].
Compared with the quadratic model in RSM, ANN has also beenused to model the process. ANN is an information processingmodel simulated from the biological neuron system. Theinterrelated network construction of direct processing basicsconducts the data processing with parallel computation [29,30].
In the present study, heterogeneous transesterification of crudejatropha oil using immobilized R. oryzae lipase catalyst has beeninvestigated. This reaction was carried-out in a batch process usingmethanol under different reaction conditions based on theexperimental design to produce biodiesel under ambient reactionconditions. In addition, RSM and ANN were applied to determinethe optimum reaction conditions.
2. Materials and methods
2.1. Materials
R. oryzae lipase and methanol were obtained from Sigma–Aldrich (USA), whereas boric acid and PVA 72000 were supplied byMerck Schuchardt OHG (Germany). Other necessary chemicals,such as calcium chloride and sodium sulphate, were procured fromGCE laboratory Chemicals (UK). Sodium alginate was obtainedfrom Fluka, Switzerland while Jatropha oil was supplied by a localcompany in Malacca, Malaysia. The fatty acid compositions of theoil and its characteristics are reported in Table 1.
2.2. Statistical methods
A five level central composite design with four factors wasemployed to determine the effect of variables, such as reactiontemperature (A), reaction time (B), methanol/oil ratio (C), andwater content (D), on FAME yield. Feasibility experiments wereconducted before the experimental design to determine the rangeof these four variables; the selected ranges are shown in Table 2.
The total number of experimental runs was calculated using theformula: 2K + 2K + 6, where K is the number of variables. Thenumber of experiments for the current study was 30, wherein 24experiments and 6 imitation tests at the design center wereconducted randomly to estimate the error.
Table 1Physical properties and chemical composition of jatropha oil.
Property Test method Results
Viscosity at 40 8C (mm2/s) ASTM D-445 34.20
Density at 15 8C (kg/m3) ASTM D-4052 908
Acid Value (mg KOH/g) ASTM D-974 3.83
Water content (%) ASTM D-2709 0.09
Fatty acid profile (%) ASTM D-1983
Palmitic acid 15.3
Stearic acid 5.8
Oleic acid 43.5
Linoleic acid 33.7
Others 1.7
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
Eq. (1) is a suitable and simple model for optimization. Itsresponse is based on selected variables by quadratic and linearterms as follows:
h ¼ bo þXk
j¼1
b jx j þXk
j¼1
b j jx2j þ
Xi
Xk
< j¼2
bi jxix j þ ei (1)
where h is the response, bo is the constant coefficient, xi and xj arethe independent factors, bj, bjj, and bij are the coefficients for linear,quadratic, and interaction effects, respectively, and ei is the error.The polynomial quality was fitted by the determination coeffi-cients, R2 and R2
ad j, in Eqs. (2) and (3) [31].
R2 ¼ 1 � SSresidual
SSmodel � SSresidual(2)
R2ad j ¼ 1 � SSresidual=DFresidual
ðSSmodel þ SSresidualÞ=ðDFmodel þ DFresidualÞ(3)
In these equations, DF is the degree of freedom and SS is the sumof squares. Subsequently, the response yield was measured at theend of the experiment. MATLAB 7.11.0 (R2010b) and Design-Expert 8.0.6 softwares were utilized for appropriate modeling,optimization, and experimental design.
2.3. Artificial neural network
The back propagation (BP) network with Marquadt Levenbergalgorithm (MLA) was used for the selected model because of its fasttraining in ANN. Fig. 1 displays that the BP network includes theinput, hidden, and output layers. The input variable values areforwarded to the hidden layers by the input nodes, followed by theestimation of the final model by the output nodes.
In ANN, several important parameters exist in network training,including the learning rate, which influences the speed of training.A moderate learning rate can help obtain a good network in a shorttime. The number of neurons in a hidden layer is another factor fora good network; network performance can be increased with moreneurons. However, a constantly high number of neurons do notindicate better prediction capability.
The data in the training process were divided into three parts,namely, training, validation, and testing. The training data wereused in the network during training; the weights and biases wereadjusted according to their error. Validation data were used tomeasure network generalization. The training can be stoppedwhen generalization stops improving. The two sets of data dictatethe performance of the approximation function and the predictioncapacity. The testing data do not affect the training, they onlyprovide an independent measure of network performance duringand after training. A training algorithm, which is a rule of ANN, wasused to update the weights and biases of the network to reduce theerror caused by the contrast between the values of the simulateddata and the target (experimental data). In the current study, MLAwas adopted to assist these parameters.
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
Fig. 1. BP network mode.
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx 3
G Model
JTICE-654; No. of Pages 8
2.4. Catalyst immobilization
The immobilization technique used was based on a previousstudy [32], but sodium sulfate was utilized as a treatment solutionat the end of the immobilization process in this study.
2.5. Surface morphologies
Field emission scanning electron microscopy was used toexamine the surface morphologies and analyze the trace elementin the PVA-alginate beads, as well as the cross-sectional structure ofthe beads. Fig. 2 clearly portrays the formed layers of the cross-linked polymers in the inner surface of the beads, where themicrostructure and stability of the beads with uniformly distributedlayers can be distinguished. The stability of the beads can beattributed to the use of highly concentrated PVA. The beads hadmore compact layers with much denser structure; thus they werenot dissolved in a warm and shaking mixture during the process[33]. The formation of the pores on the outer layer of the beads isrevealed in Fig. 3. The pores were uniformly distributed withaverage sizes. Moreover, the surface morphology of the supportdisplayed that the enzyme was attached to the support layers.
3. Experimental
3.1. Catalyst preparation
A specific amount of lipase was mixed with phosphate andstirred at 140 rpm for 10 min. The mixture was then kept at 4 8C
Fig. 2. Image of lipase immobilized in PVA-Alginate beads taken using FESEM.
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
before filtration under sterile conditions. PVA and sodium alginatewere utilized for enzyme immobilization. After preparing the PVAsolution, sodium alginate and lipase were blended into thesolution. The mixture was then injected drop wise into a solutioncontaining boric acid and calcium chloride, with continuousstirring for 30–50 min to produce the beads. The beads were keptat 4 8C for 24 h. Subsequently, 10% boric acid solution was addedwhile the beads were stirred for 30 min and treated with 0.5 Msodium sulphate solution for another 30 min. The beads were thenstored at 4 8C for further use. The experiment was conducted understerile conditions.
3.2. Biodiesel production
Biodiesel was produced using a 2 mm thick glass vial as thereactor vessel. The vial was sealed tightly with a cap of silicon rubberto preserve the vaporized compositions. The reactor vessel wasplaced in an incubator shaker for agitation and warming. Thereaction mixture consisted of crude jatropha oil, immobilized lipase,methanol, and water. At a certain reaction interval, a sample reactionmixture was extracted and centrifuged. A 100 mL sample was thenwithdrawn from the supernatant and analyzed for biodiesel yieldusing gas chromatograph–mass spectroscopy (GC–MS). Each testwas carried out in triplicate to ensure reproducibility.
3.3. Sample analysis
The major properties of biodiesel, such as kinematic viscosity,neutralization value, flash point, cloud point, pour point, anddensity, obtained according to the American Society for Testing
Fig. 3. Outer layer of a bead.
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
Table 3Fuel properties of obtained biodiesel from jatropha oil.
Property Test method ASTM
specification
Biodiesel
Viscosity at 40 8C (mm2/s) ASTMD-445 1.9 to 6.0 4.39
Density at 15 8C (kg/m3) ASTM D-4052 875 to 900 876
Neutralization value
(mg KOH/g)
ASTM D-664 Max 0.5 0.27
Cloud point (8C) ASTM D-2500 �3 to 12 4
Pour point (8C) ASTM D-97 �15 to 10 5
Flash point (8C) ASTM D-93 Min 130 163
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx4
G Model
JTICE-654; No. of Pages 8
and Materials International (ASTM) D6751 method are listed inTable 3. The biodiesel product was analyzed using AgilentTechnologies 6890 N GC–MS with an inert mass selective detector5975. The capillary column was an Agilent 19091S-433 HP-5MS(30 mm � 250 mm � 0.25 mm), with helium as the carrier gas. Theinitial temperature of the oven was 80 8C for 30 s, and thensubsequently increased at the rate of 10 8C/min until thetemperature reached 250 8C, which was held constant for 5 min.The analysis was conducted for a total period of 42 min. Thetemperature of the detector was 250 8C, whereas the temperatureof the injector was 325 8C. FAME yield was calculated using Eq. (4).
Yield% ¼ ðFAME area from GCMS � weight of productÞðweight of oil sampleÞ � 100 (4)
4. Results and discussion
4.1. Experimental results
The relationships between the four independent variables(reaction temperature, time of reaction, methanol/oil ratio, and
Table 4Design of experiments and RSM, ANN and experimental results.
Run Temperature (8C) Reaction time (h) Methanol/oil ratio
(mol/mol)
W
1 35.00 5.00 4.00
2 45.00 5.00 4.00
3 35.00 13.00 4.00
4 45.00 13.00 4.00
5 35.00 5.00 6.00
6 45.00 5.00 6.00
7 35.00 13.00 6.00
8 45.00 13.00 6.00
9 35.00 5.00 4.00 1
10 45.00 5.00 4.00 1
11 35.00 13.00 4.00 1
12 45.00 13.00 4.00 1
13 35.00 5.00 6.00 1
14 45.00 5.00 6.00 1
15 35.00 13.00 6.00 1
16 45.00 13.00 6.00 1
17 30.00 9.00 5.00
18 50.00 9.00 5.00
19 40.00 1.00 5.00
20 40.00 17.00 5.00
21 40.00 9.00 3.00
22 40.00 9.00 7.00
23 40.00 9.00 5.00
24 40.00 9.00 5.00 1
25 40.00 9.00 5.00
26 40.00 9.00 5.00
27 40.00 9.00 5.00
28 40.00 9.00 5.00
29 40.00 9.00 5.00
30 40.00 9.00 5.00
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
water content) and biodiesel yield were studied. The FAME yieldfor each experimental run and the yields from both ANN and RSMmodels are listed in Table 4.
4.2. Regression model and statistical analysis by RSM
The RSM response obtained in Table 4 corresponded to the fourindependent variables utilizing a polynomial equation such as Eq.(1). Least squares regression was employed to fit the acquired datato Eq. (1). The best fit model in terms of actual factors for the FAMEyield is described in Eq. (5).
Yield ¼ �20:83 þ 3:62A þ 1:22B þ 6:36C þ 0:22D þ 0:01AB
� 0:20AC � 2:3 � 10�3AD þ 0:40BC � 2:86 � 10�3BD
þ 9:37 � 10�3CD � 0:04A2 � 0:06B2 � 0:70C2 � 1:05
� 10�3D2 (5)
The ANOVA assessments of the best fit model indicated that themodel is suitable for describing the experimental work, as shownin Table 5. The F-value at 51.88 indicated the significance of thequadratic model. Furthermore, the effect of each term in the modelwas evaluated; the highly significant effect of some variables wasconcluded based on the F-value. The linear terms for reactiontemperature (A), time (B), methanol/oil ratio (C), and all of thevariables in quadratic form, as well as the interaction terms for twofactors (AD and BD) tabulated in Table 5 have significantly affectedthe biodiesel yield because of the high F-values and low p-values(<0.05).
A p-value of greater than 0.05 indicates that the correspondingmodel term is not significant. Table 6 shows that the twointeraction terms, namely, temperature–methanol/oil ratio (AC)and time–methanol/oil ratio (BC), registered higher p-values. Thus,
ater content (wt%) Exp. yield (%) ANN yield (%) RSM yield (%)
40.00 77.50 77.99 77.83
40.00 79.00 78.88 78.77
40.00 84.00 84.20 83.45
40.00 85.00 86.42 85.51
40.00 76.00 76.17 76.43
40.00 77.00 76.84 77.00
40.00 82.00 81.79 82.68
40.00 84.00 85.13 84.37
00.00 79.00 80.24 78.93
00.00 79.00 79.15 78.50
00.00 83.00 83.46 83.18
00.00 84.00 83.93 83.87
00.00 79.00 79.06 78.66
00.00 77.00 77.14 77.85
00.00 83.00 83.45 83.53
00.00 84.00 83.89 83.85
70.00 80.00 80.90 79.64
70.00 81.00 81.18 80.89
70.00 75.00 75.70 75.00
70.00 87.10 87.07 86.62
70.00 82.00 82.34 82.47
70.00 82.00 81.70 81.05
10.00 81.00 81.25 80.47
30.00 81.00 81.05 81.05
70.00 84.00 84.60 84.55
70.00 85.00 84.60 84.55
70.00 84.50 84.60 84.55
70.00 85.00 84.60 84.55
70.00 84.00 84.60 84.55
70.00 84.80 84.60 84.55
MSE 0.28 0.19
R2 0.97 0.98
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
Table 5Analysis of variance (ANOVA) for regression model.
Source Sum of squares df Mean Square F value p-Value prob > F
Model 281.4305 14 20.10218 51.87659 < 0.0001 Significant
A-Temp. 2.34375 1 2.34375 6.048387 0.0265
B-time 202.4204 1 202.4204 522.3753 < 0.0001
C-ratio 3.010417 1 3.010417 7.768817 0.0138
D-Water cont. 0.510417 1 0.510417 1.317204 0.2691
AB 1.265625 1 1.265625 3.266129 0.0908
AC 0.140625 1 0.140625 0.362903 0.5559
AD 1.890625 1 1.890625 4.879032 0.0432
BC 0.390625 1 0.390625 1.008065 0.3313
BD 1.890625 1 1.890625 4.879032 0.0432
CD 1.265625 1 1.265625 3.266129 0.0908
A^2 31.51313 1 31.51313 81.32419 < 0.0001
B^2 23.9467 1 23.9467 61.79793 < 0.0001
C^2 13.32027 1 13.32027 34.37488 < 0.0001
D^2 24.5917 1 24.5917 63.46244 < 0.0001
Residual 5.8125 15 0.3875
Lack of Fit 4.7375 10 0.47375 2.203488 0.1983 Not significant
Pure Error 1.075 5 0.215
Cor Total 287.243 29
Table 6ANOVA for response surface reduced quadratic model.
Source Sum of squares df Mean square F value p-Value prob > F
Model 280.8993 12 23.40827 62.72955 <0.0001 Significant
A-Temperature 2.34375 1 2.34375 6.280788 0.0227
B-time 202.4204 1 202.4204 542.4468 <0.0001
C-ratio 3.010417 1 3.010417 8.067323 0.0113
D-Water content 0.510417 1 0.510417 1.367816 0.2583
AB 1.265625 1 1.265625 3.391626 0.0830
AD 1.890625 1 1.890625 5.066502 0.0379
BD 1.890625 1 1.890625 5.066502 0.0379
CD 1.265625 1 1.265625 3.391626 0.0830
A^2 31.51313 1 31.51313 84.44897 <0.0001
B^2 23.9467 1 23.9467 64.17243 <0.0001
C^2 13.32027 1 13.32027 35.69569 <0.0001
D^2 24.5917 1 24.5917 65.9009 <0.0001
Residual 6.34375 17 0.373162
Lack of Fit 5.26875 12 0.439063 2.042151 0.2220 Not significant
Pure Error 1.075 5 0.215
Cor Total 287.243 29
Fig. 4. The Predicted vs. Actual plot.
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx 5
G Model
JTICE-654; No. of Pages 8
they were considered insignificant terms and were removed toimprove the current model and produce an enhanced new model.The determination coefficient (R2) and the regression equationwere evaluated to test the fit of the model [34]. The R2 of thereduced model decreased from 0.9798 to 0.9779 because ofvariable reduction from the regression model [35]. Meanwhile, theadjusted determination coefficient (Radj) increased from 0.9609 to0.9623 after excluding unnecessary terms [36]. The F-value of themodified model increased to 62.73, but its p-value remained<0.0001, indicating that the model improved after deleting the ACand BC terms.
The final practical model after modification based on theANOVA data and statistical parameters is given by Eq. (6) in termsof uncoded factors.
Yield ¼ �18:84 þ 3:53A þ 1:42B þ 5:96C þ 0:22D þ 0:014AB
� 2:30 � 10�3AD � 2:86 � 10�3BD þ 9:38 � 10�3CD
� 0:04A2 � 0:06B2 � 0:70C2 � 1:05 � 10�3D2 (6)
A p-value <0.0001 indicates that there is only a 0.01%probability that a model with a similarly large F-value becomesthe noise product in the data. Moreover, the lack of fit F-value was2.04, denoting that the lack of fit was not significant in pure error[37].
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
Fig. 4 plots the predicted versus the actual FAME yields andshows that the experimental and predicted values are in proximityto each other. This result indicates the accuracy of the model tocapture the correlation between biodiesel yield and the indepen-dent reaction variables.
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
Fig. 5. The effect of reaction temperature and reaction time on FAME yield, for water
content of 70 wt% and Methanol/oil Ratio of 5%.
Table 7Weights and biases of BP network.
W1a W2b
X1 X2 X3 X4 Biasa 1 �0.2295
1 5.4469 �0.1077 �0.1834 5.6632 �2.7542 2 0.2752
2 �2.5956 3.3872 �2.5042 10.8558 3.5282 3 0.1914
3 3.1446 4.1704 0.8810 �1.7310 �1.8694 4 0.0915
4 1.0327 �6.6608 �1.1976 1.2791 �1.9580 5 �0.6393
5 �0.9082 �1.2714 0.0945 0.2958 �0.5990 6 0.2344
6 �3.9774 4.2756 1.1401 �2.6631 2.1926 Biasb �0.4413
a Weights and biases between the input layer and hidden layer.b Weights and biases between the hidden layer and output layer.
Fig. 6. The effects of methanol/oil ratio and water content on FAME yield for a total
time of 17 h and reaction temperature of 40 8C.
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx6
G Model
JTICE-654; No. of Pages 8
4.3. Prediction with ANN
The weights and biases of the BP network are shown in Table 7.Biodiesel yield under different ranges of input conditions can bepredicted using this table.
The bias and weight values of the hidden, input, and outputlayer neurons were detected based on the maximum amount of R2
and on another important factor, namely, the mean squared error(MSE), which refers to the lower value for better performance, tooptimize the network parameters. These two factors werecalculated using Eqs. (7) and (8) [29]:
MSE ¼ 1
n
Xn
i¼1
y pi � yai
� �2(7)
R2 ¼ 1 �Pn
i¼1 ðy pi � yaiÞ2
Pni¼1 ðy pi � ymÞ
2(8)
where n is the data set number, yai and ypi are the actual and thepredicted output values of the ith set, respectively, and ym is theactual output value mean.
The low MSE (0.28) and high R2 (0.97) given in Table 4 indicatedthe predicted and the actual values of the experimental data fittedwell [38]. However, the slight differences signified that RSMperformance was better than ANN. Nevertheless, the results weregenerally similar and both models are acceptable.
4.4. Influence of reaction temperature and time
The statistical analysis of the experimental data indicated thattemperature (A) and time (B) were important and effectivevariables in response analysis and they have large, linear, andpositive influence on FAME yield. Fig. 5 demonstrates the surfaceplot of the response for the production of biodiesel. The biodieselyield was enhanced with the increase in reaction temperatureand then reduction trend in the yield of biodiesel can be seenafter the optimum temperature under conditions that themethanol/oil molar ratio was 5:1 and water content was 70%(w/w). Similar trend was observed for the entire range of reactiontime. Fig. 5 exhibits that the maximum biodiesel yield obtainedwas 87.1% at temperature of 8C and for 17 h reaction time. Theseconditions are reasonable compared with other reports onbiodiesel production from jatropha oil. Hawash et al. [39]produced biodiesel from jatropha oil under supercritical condi-tion, with 100% yield under 4 min reaction time at 320 8C, with8.4 MPa pressure and 43:1 methanol/oil ratio. These conditionsled to costly biodiesel production process. Tamalampudi et al.
[26] produced biodiesel from jatropha oil using R. oryzae onpolyurethane foam particles with 80% yield under ambientconditions (T = 30 8C and alcohol/oil ratio = 3:1), but with longerreaction time (60 h).
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
4.5. Methanol/oil ratio and water content effects on FAME yield
Based on the ANOVA results, the methanol/oil molar ratio isconsidered as another important parameter that affects thebiodiesel yield. The effect of methanol/oil ratio (C) and moisturecontent (D) on biodiesel yield under the condition where thereaction temperature and time were constant at 40 8C and 17 h,respectively, is shown in Fig. 6. The maximum FAME yield was87.1% in a methanol/oil molar ratio of 5:1 and water content of70 wt%. The biodiesel yield decreased with the alcohol/oil ratiobeyond 5:1.
Based on the stoichiometry of the transesterification reaction,at least three moles of alcohol are required to produce three molesof ester for a reaction with one mole of oil. The reversibility of thereactions and the methanol/oil ratio produce more methyl esterproducts. However, the excess amount of methanol hindersglycerol recovery because of the effect of methanol on glycerolsolubility [40–42]. The existence of glycerol in the solution drivesthe equilibrium back to the left hand side and consequently lowersthe biodiesel yield [42]. Furthermore, the high amount of methanol
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx 7
G Model
JTICE-654; No. of Pages 8
negatively affects the lipase catalyst activity. Thus, the yieldreduced after the alcohol/oil ratio exceeded the optimum amount.
Water content is another important factor in biocatalytictransesterification reaction by helping to avoid lipase deactivation.The biodiesel yield in Fig. 6 increased with the enhancement inwater content but decreased after the water content reached acertain value. The reduction of the transesterification yield can beattributed to hydrolysis as have been reported in previous studies[43–45].
4.6. Validation of the model
The following optimal reaction conditions: temperature (40 8C),time (17 h), methanol/oil molar ratio (5:1), and water content(70 wt%) were determined using the regression model. Triplicateexperiments were carried out at these optimum conditions toconfirm the accuracy and validation of the regression model. Theaverage yield of biodiesel obtained at 87.10% was close to thepredicted value of 87.07%.
5. Conclusions
The experimental results indicated that 87.1% of biodiesel yieldwas obtained under the reaction conditions of 40 8C temperaturewithin 17 h reaction time, 5:1 alcohol/oil molar ratio, and 70 wt%water content. The predicted FAME yield was experimentallyvalidated. Both the predicted and experimental results fitted well,where the R2 for RSM and ANN models were 0.98 and 0.97,respectively. The models can provide good alternatives tocumbersome laboratory testing. The immobilized lipase catalystexhibited good activity for biodiesel production, as indicated bythe major properties that complied with the ASTM D6751standard. Therefore, immobilized R. oryzae lipase has highpotential to be used in the enzymatic production of biodieselunder ambient reaction conditions.
Acknowledgements
The authors would like to express their sincere gratitude to theMinistry of Higher Education (MOHE), Malaysia for supporting theproject under the Fundamental Research Grant Scheme (FRGS)Vots 78402 and 78681.
References
[1] Zarei A, Amin NAS, Talebian-Kiakalaieh A, Jalilian H. Biodiesel production withimmobilized enzyme and effect of alcohol on the result. IACSIT Press 2012;38:141–4.
[2] Shang N-C, Liu R-Z, Chen Y-H, Chang C-Y, Lin R-H. Characterization of fatty acidmethyl esters in biodiesel using high-performance liquid chromatography. JTaiwan Inst Chem Eng 2012;43:354–9.
[3] Chen L-Y, Chen Y-H, Hung Y-S, Chiang T-H, Tsai C-H. Fuel properties andcombustion characteristics of jatropha oil biodiesel–diesel blends. J TaiwanInst Chem Eng 2013;44:214–20.
[4] Talebian-Kiakalaieh A, Amin NAS, Zarei A, Noshadi I. Transesterification ofwaste cooking oil by heteropoly acid (HPA) catalyst: optimization and kineticmodel. Appl Energy 2013;102:283–92.
[5] Chen Y-H, Chen J-H, Luo Y-M, Shang N-C, Chang C-H, Chang C-Y, et al. Propertymodification of jatropha oil biodiesel by blending with other biodiesels oradding antioxidants. Energy 2011;36:4415–21.
[6] Ting W-J, Huang C-M, Giridhar N, Wu W-T. An enzymatic/acid-catalyzedhybrid process for biodiesel production from soybean oil. J Chin Inst ChemEng 2008;39:203–10.
[7] Wang J-X, Chen K-T, Wen B-Z, Liao Y-HB, Chen C-C. Transesterification ofsoybean oil to biodiesel using cement as a solid base catalyst. J Taiwan InstChem Eng 2012;43:215–9.
[8] Shahbazi MR, Khoshandam B, Nasiri M, Ghazvini M. Biodiesel production viaalkali-catalyzed transesterification of Malaysian RBD palm oil – Characteriza-tion, kinetics model. J Taiwan Inst Chem Eng 2012;43:504–10.
[9] Antolin G, Tinaut FV, Briceno Y, Castano V, Perez C, Ramirez AI. Optimisation ofbiodiesel production by sunflower oil transesterification. Bioresour Technol2002;83:111–4.
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
[10] Jiang J-J, Tan C-S. Biodiesel production from coconut oil in supercriticalmethanol in the presence of cosolvent. J Taiwan Inst Chem Eng 2012;43:102–7.
[11] Agarwal M, Chauhan G, Chaurasia SP, Singh K. Study of catalytic behavior ofKOH as homogeneous and heterogeneous catalyst for biodiesel production. JTaiwan Inst Chem Eng 2012;43:89–94.
[12] Huang G, Chen F, Wei D, Zhang X, Chen G. Biodiesel production by microalgalbiotechnology. Appl Energy 2010;87:38–46.
[13] Kaewcharoensombat U, Prommetta K, Srinophakun T. Life cycle assessmentof biodiesel production from jatropha. J Taiwan Inst Chem Eng 2011;42:454–62.
[14] Naik M, Meher LC, Naik SN, Das LM. Production of biodiesel from high free fattyacid Karanja (Pongamia pinnata) oil. Biomass Bioenergy 2008;32:354–7.
[15] Yang R, Su M, Zhang J, Jin F, Zha C, Li M, et al. Biodiesel production from rubberseed oil using poly (sodium acrylate) supporting NaOH as a water-resistantcatalyst. Bioresour Technol 2011;102:2665–71.
[16] Sahoo PK, Das LM. Combustion analysis of Jatropha, Karanja and Polanga basedbiodiesel as fuel in a diesel engine. Fuel 2009;88:994–9.
[17] Tan KT, Lee KT, Mohamed AR. Potential of waste palm cooking oil for catalyst-free biodiesel production. Energy 2011;36:2085–8.
[18] Chen Y-H, Chiang T-H, Chen J-H. An optimum biodiesel combination: Jatrophaand soapnut oil biodiesel blends. Fuel 2012;92:377–80.
[19] Jain S, Sharma MP. Prospects of biodiesel from Jatropha in India: a review.Renew Sustain Energy Rev 2010;14:763–71.
[20] Basha SA, Gopal KR, Jebaraj S. A review on biodiesel production, combus-tion, emissions and performance. Renew Sustain Energy Rev 2009;13:1628–34.
[21] Lee HV, Yunus R, Juan JC, Taufiq-Yap YH. Process optimization design forjatropha-based biodiesel production using response surface methodology.Fuel Process Technol 2011;92:2420–8.
[22] Salamatinia B, Kamaruddin AH, Abdullah AZ. Regeneration and reuse of spentNaOH-treated oil palm frond for copper and zinc removal from wastewater.Chem Eng J 2010;156:141–5.
[23] Emregul E, Sungur S, Akbulut U. Polyacrylamide–gelatine carrier system usedfor invertase immobilization. Food Chem 2006;97:591–7.
[24] Danisman T, Tan S, Kacar Y, Ergene A. Covalent immobilization of invertase onmicroporous pHEMA–GMA membrane. Food Chem 2004;85:461–6.
[25] Leuschner F, inventor German Patent NO. 1,227,855 1996.[26] Tamalampudi S, Talukder MR, Hama S, Numata T, Kondo A, Fukuda H.
Enzymatic production of biodiesel from Jatropha oil: a comparative studyof immobilized-whole cell and commercial lipases as a biocatalyst. BiochemEng J 2008;39:185–9.
[27] Charoenchaitrakool M, Thienmethangkoon J. Statistical optimization for bio-diesel production from waste frying oil through two-step catalyzed process.Fuel Process Technol 2011;92:112–8.
[28] Noshadi I, Amin NAS, Parnas RS. Continuous production of biodiesel fromwaste cooking oil in a reactive distillation column catalyzed by solid hetero-polyacid: optimization using response surface methodology (RSM). Fuel 2012;94:156–64.
[29] Basri M, Rahman RN, Ebrahimpour A, Salleh AB, Gunawan ER, Rahman MBA.Comparison of estimation capabilities of response surface methodology (RSM)with artificial neural network (ANN) in lipase-catalyzed synthesis of palm-based wax ester. BMC Biotechnol 2007;7:53.
[30] Yuste AJ, Dorado MP. A neural network approach to simulate biodieselproduction from waste olive oil. Energy Fuels 2006;20:399–402.
[31] Ozer A, Gurbuz G, Calimli A, Korbahti BK. Biosorption of copper(II) ions onEnteromorpha prolifera: application of response surface methodology (RSM).Chem Eng J 2009;146:377–87.
[32] Idris A, Zain NAM, Suhaimi MS. Immobilization of Baker’s yeast invertase inPVA–alginate matrix using innovative immobilization technique. Proc Bio-chem 2008;43:331–8.
[33] Zain NAM, Suhaimi MS, Idris A. Development and modification ofPVA–alginate as a suitable immobilization matrix. Proc Biochem 2011;46:2122–9.
[34] Yuan X, Liu J, Zeng G, Shi J, Tong J, Huang G. Optimization of conversion ofwaste rapeseed oil with high FFA to biodiesel using response surface meth-odology. Renew Energy 2008;33:1678–84.
[35] Mazaheri H, Lee KT, Bhatia S, Mohamed AR. Subcritical water liquefaction of oilpalm fruit press fiber in the presence of sodium hydroxide: an optimisationstudy using response surface methodology. Bioresour Technol 2010;101:9335–41.
[36] Fermoso J, Gil MV, Arias B, Plaza MG, Pevida C, Pis JJ, et al. Application ofresponse surface methodology to assess the combined effect of operatingvariables on high-pressure coal gasification for H2-rich gas production. Int JHydrogen Energy 2010;35:1191–204.
[37] Wu J, Wang JL, Li MH, Lin JP, Wei DZ. Optimization of immobilization forselective oxidation of benzyl alcohol by Gluconobacter oxydans using re-sponse surface methodology. Bioresour Technol 2010;101:8936–41.
[38] Rajendra M, Jena PC, Raheman H. Prediction of optimized pretreatmentprocess parameters for biodiesel production using ANN and GA. Fuel 2009;88:868–75.
[39] Hawash S, Kamal N, Zaher F, Kenawi O, Diwani GE. Biodiesel fuel from Jatrophaoil via non-catalytic supercritical methanol transesterification. Fuel 2009;88:579–82.
[40] Ma F, Hanna MA. Biodiesel production: a review. Bioresour Technol 1999;70:1–15.
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015
A. Zarei et al. / Journal of the Taiwan Institute of Chemical Engineers xxx (2013) xxx–xxx8
G Model
JTICE-654; No. of Pages 8
[41] Noureddini H, Zhu D. Kinetics of transesterification of soybean oil. J Am OilChem Soc 1997;74:1457–63.
[42] Silva GF, Camargo FL, Ferreira ALO. Application of response surface method-ology for optimization of biodiesel production by transesterification of soy-bean oil with ethanol. Fuel Process Technol 2011;92:407–13.
[43] Dizge N, Keskinler B. Enzymatic production of biodiesel from canola oil usingimmobilized lipase. Biomass Bioenergy 2008;32:1274–8.
Please cite this article in press as: Zarei, A., et al., Immobilized lipaseand modeling. J. Taiwan Inst. Chem. Eng. (2013), http://dx.doi.org/10
[44] Li X, He X-Y, Li Z-L, Wang Y-D, Wang C-Y, Shi H, et al. Enzymatic production ofbiodiesel from Pistacia chinensis bge seed oil using immobilized lipase. Fuel2012;92:89–93.
[45] Noureddini H, Gao X, Philkana RS. Immobilized Pseudomonas cepacia lipasefor biodiesel fuel production from soybean oil. Bioresour Technol 2005;96:769–77.
-catalyzed transesterification of Jatropha curcas oil: Optimization.1016/j.jtice.2013.05.015