Article pubs.acs.org/EF
Biodiesel Production from Soybean Oil in a Membrane Reactor over Hydrotalcite Based Catalyst: An Optimization Study Wei Xu, Lijing Gao, Songcheng Wang, and Guomin Xiao* School of Chemistry and Chemical Engineering, Southeast University, Nanjing, 211189, P. R. China ABSTRACT: Transesterification of soybean oil to produce biodiesel was performed in a membrane reactor packed with shaped KF/Ca−Mg−Al hydrotalcite solid base. The microfiltration ceramic membrane (length: 20 cm, inner/outer diameter: 6/10 mm) was used to retain the oil during the transesterification reaction. High quality biodiesel was produced in the fixed bed membrane reactor by coupling heterogeneous alkali catalyzed transesterification and separation process. The response surface methodology (RSM) was employed to evaluate the effects of such factors as reaction temperature, catalyst amount, and circulation velocity on biodiesel production. 70 °C reaction temperature, 0.531 g/cm3 catalyst amount and 3.16 mL/min circulation velocity was found to be the optimum condition, achieving a 0.1820 g/min biodiesel yielding rate during 150 min of circulation time.
1. INTRODUCTION Rising petroleum prices and increasing environmental concern have attracted international attention to develop alternative nonpetroleum fuels. Biodiesel, a mixture of mono alkyl (normally methyl) ester of long chain fatty acids, is a clean and renewable fuel with less impact on the environment. The most common biodiesel conversion technology is the transesterification between vegetable oil and methanol, including catalytic or noncatalytic processes.1,2 Both the catalytic and noncatalytic transesterification downstream processes will obtain a mixture of biodiesel and glycerol, as well as the remaining reactant and catalyst.3 Currently, the removal of residual triglyceride and glycerol from the biodiesel product is one of the main factors that plague biodiesel production costs apart from the cost of raw materials.4 Membrane reaction process, as a promising intensification technology, which combines reaction and separation in a single unit, can overcome the equilibrium limitation by continuous removal of the products. Since the transesterification is an equilibrium reaction and downstream processing is a particular concern, the membrane reactor offers an efficient and practical technology for biodiesel production.5,6 Furthermore, the membrane reactor also provides sufficient mixing of reactants and obtains high masstransfer between the immiscible phases. It is known that conventional fatty acid methyl ester (FAME) is entirely miscible in methanol at normal reaction temperatures,7,8 whereas the menthol and vegetable oil phases are immiscible for all practical purposes. The formation of a twophase (emulsified) system provides the possibility of the membrane reaction operation. Dube et al.9 presented the principle of membrane reactor operation: the oil exists in the form of an emulsion due to the immiscibility and the various surface forces, so the oil droplets cannot pass though the pores of the membrane; whereas the FAME was soluble in methanol, they will pass though the membrane along with glycerol. Based on this principle, applications of membrane reactor for biodiesel production from different vegetable oil have been reported recently. For both acid/base-catalysis, membrane reactor was particularly useful in removing unreacted oil from the FAME and shifting the reaction equilibrium to the product side.8−12 © 2013 American Chemical Society
The researchers indicated that high purity FAME can be produced in membrane reactors with lower costs and microfiltration membranes were suitable to retain the oil in the reactor.7 However, most of the catalysts used in these studies were homogeneous ones (NaOH12,13 or H2SO49), which cannot avoid the inherent flaws of homogeneous catalysis. This will greatly limit the environmental benefits of membrane separation technology. Moreover, the catalysts can dissolve in the permeate stream, which will directly affect the quality of biodiesel. A highly efficient and stable heterogeneous catalyst could not only overcome these problems but also prevent the decrease in catalytic activity in a membrane reactor found by Baroutian et al., which was caused by the formation of soap,14 thus providing a greener way to produce biodiesel. In this work, heterogeneous alkali was used in a membrane reactor in order to obtain a process in which catalyzed transesterification and membrane separation processes were coupled. Shaped KF/Ca−Mg−Al hydrotalcite catalyst, which was found very active and stable in transesterifications,15,16 was used and response surface methodology (RSM) coupled with central composite design (CCD) was utilized to analyze and optimize the process.
2. EXPERIMENTAL SECTION 2.1. Chemicals and Catalyst Preparation. The raw soybean oil (food grade) was purchased from the local market. CH3OH (AR), KF· 2H2O (AR), Ca(NO3)2·4H2O (AR), Mg(NO3)2·6H2O (AR), Al(NO3)3·9H2O (AR), NaOH (AR), and Na2CO3 (AR) were purchased from Sinopharm Chemical Reagent Co., Ltd. Ceramic membrane was manufactured by Foshan Ceramics Research Institute & Jingang Group. The length, inner diameter, outer diameter, and pore size of ceramic membrane cylinder were 200 mm, 6 mm, 10 mm, and 0.2 μm respectively. The shaped KF/Ca−Mg−Al hydrotalcite catalyst used in the study was prepared according to the procedure described in our previous work.15,16 The length and diameter of the cylinder catalyst were both 1 mm. Received: July 2, 2013 Revised: October 29, 2013 Published: October 30, 2013 6738
dx.doi.org/10.1021/ef401823z | Energy Fuels 2013, 27, 6738−6742
Energy & Fuels
Article
Figure 1. Schematic diagram of membrane reactor for biodiesel production. 2.2. Transesterification in the Membrane Reactor. A laboratory scale membrane reactor for biodiesel production was showed as the schematic diagram in Figure 1. According to the previous studies,17 a relatively higher ratios of methanol/oil was essential for the reaction system to maintain the two-phase system so that the oil-rich phase can be retained by the membrane. So, in this study, soybean oil (30 g) and methanol were initially mixed in the mixing vessel with a molar ratio of 1:24. Various amounts of catalyst were mixed with spring type packing and packed into the ceramic membrane tube. The preheated reactant mixture was continuously pumped into the reactor at a certain temperature. The transmembrane pressure was controlled by the back pressure controller and maintained 50 kPa in this study. The permeate stream was collected in a flask equipped with methanol distillation unit, in which methanol was distilled and then returned to the feed flask. After a period of time (typically 150 min), the products in the flask were transferred into a separating funnel to separate biodiesel from glycerol. The mass of the products was measured to calculate the yielding rate of FAME. The system was fully drained after each run, and then flushed for 10 min with methanol. The contents of FAME in the samples were analyzed by the gas chromatograph (Ouhua GC 9160) equipped with a DB-5Ht capillary column (15 m × 0.25 mm × 0.25 mm) equipped with a flame ionization detector. Nitrogen was used as the carrier gas and the injector and detector temperature were both 360 °C. 2.3. Experimental Design. In this work, Design Expert software Version 7.0 (Stat-Ease Inc., Minneapolis, MN) was used to design the experiments and optimize the biodiesel production. Experiments were designed by RSM coupled with CCD. Reaction temperature, catalyst amount (catalyst per unit volume of reactor) and circulation velocity of the reactant were selected as the independent parameters. The response was the yielding rate of biodiesel within 150 min. The number of replicates was chosen as six to provide a relatively stable standard error of prediction. As was shown in Table 1, the actual levels of the parameters were 50− 70 °C reaction temperature, 0.177−0.531 g/cm3 catalyst amount and 2.8−3.8 mL/min circulation velocity. The coded values were designated by −1, 0, +1, -alpha, and +alpha. In order to create a face-centered central composite design, alpha was set to 1 in this work. This was desirable because it was a three level design, and this also ensured that the axial runs will be less extreme values than the factorial portion.18 Selection of levels of reaction temperature was based on the conventional reaction temperature of transesterifications using KF/Ca− Mg−Al hydrotalcite catalyst. The levels of circulation velocity were selected according to the capacity of the circulation pump and the catalytic efficient of the catalyst. A relatively high circulation velocity (low resistance time) was selected because of the phase inversion limitations that the single pass conversion cannot be that high.
Table 1. Experimental Design Matrix and the Response Value reaction temperature (°C)
catalyst amount (g/cm3)
circulation velocity (mL/min)
run
actual
coded
actual
coded
actual
coded
yielding rate (g/min)
19 15 17 20 12 2 10 7 14 9 4 6 13 16 5 3 8 1 11 18
60 60 60 60 60 70 70 50 60 50 70 70 60 60 50 50 70 50 60 60
0 0 0 0 0 1 1 −1 0 −1 1 1 0 0 −1 −1 1 −1 0 0
0.354 0.354 0.354 0.354 0.534 0.177 0.354 0.534 0.354 0.354 0.534 0.177 0.354 0.354 0.177 0.534 0.534 0.177 0.177 0.354
0 0 0 0 1 −1 0 1 0 0 1 −1 0 0 0 1 1 −1 −1 0
3.3 3.3 3.3 3.3 3.3 2.8 3.3 3.8 3.8 3.3 2.8 3.8 2.8 3.3 3.8 2.8 3.8 2.8 3.3 3.3
0 0 0 0 0 −1 0 1 1 0 −1 1 −1 0 1 −1 1 −1 0 0
0.1392 0.1421 0.1415 0.1413 0.1524 0.1208 0.1615 0.0845 0.1118 0.1148 0.1682 0.1002 0.1283 0.1404 0.0756 0.1087 0.1456 0.0924 0.1179 0.1428
3. RESULTS AND DISCUSSION 3.1. Response Surface Methodology. The results of the experimental design matrix and the yielding rate of FAME (response value) are shown in Table 1. A quadratic model is fitted to the experimental results assisted by the Design Expert software. Diagnostics of the residuals indicate that transformation is not required to improve the model. Equations 1 and 2 show the quadratic models to predict the FAME yielding rate in terms of coded and actual factors, where T, M, and V represents the reaction temperature (°C), catalyst amount (g/cm3), and circulation velocity (mL/min), respectively. FAME yielding rate(g/min) = +0.14 + 0.022 × T + 0.015 × M − 0.010 × V + 0.00845 × T × M − 0.000275 × T × V − 0.001175 × M × V − 0.002673 × T 2 − 0.005673 × M2 − 0.021 × V 2 6739
(1)
dx.doi.org/10.1021/ef401823z | Energy Fuels 2013, 27, 6738−6742
Energy & Fuels
Article
that model terms are significant. Values greater than 0.1 indicate the model terms are not significant. The ANOVA table shows that the reaction temperature, catalyst amount, circulation velocity, the quadratic terms of these three variables and the interaction of reaction temperature and catalyst amount are significant. The p-value of “lack of fit” is greater than 0.1, indicating that the model fits to all data well. The predicted R2 of 0.9778 reasonably agreed with the adjusted R2 of 0.9946. The signal-tonoise ratio is measured by adequate precision and a ratio greater than 4 is desirable. This model can be used to navigate the design space as the signal-to-noise ratio of 71.668 indicates an adequate signal. Due to the high p-values, the interactions of reaction temperature-circulation velocity and circulation velocity-catalyst amount are found not significant thus the removal of these terms seems very tempting. However, since the value is between the critical values of significant and insignificant and the removal of these terms will decrease the p-value of “lack of fit” below 0.1, they cannot be removed. 3.2. Effects of Process Parameters. Based on the model analysis, it can be observed that the reaction temperature has the greatest effect on the yielding rate, and then followed the catalyst amount and the circulation velocity according to the F-value. Moreover, reaction temperature and catalyst amount have positive effect on the FAME yielding rate, whereas the circulation velocity has negative effect on it according to eq 1. Three-dimensional surface plots of the predicted FAME yielding rate are shown in Figure 3 based on eq 2. As presented in Figure 3(a), the FAME yielding rate increases significantly with the catalyst amount at higher temperatures because of the positive effect of interaction between catalyst amount and reaction temperature (TM in eq 1). Figure 3(b) and (c) show the relationship between FAME yielding rate and the interaction between catalyst amount and circulation velocity, circulation velocity and reaction temperature, respectively. As can be seen in these figures, the yielding rate of FAME increases with the rising of reaction temperature and the catalyst amount. This is in agreement with the facts reported in literatures that the rising of reaction temperature and catalyst amount could increase the yield of FAME in a fixed bed reactor using heterogeneous catalyst. As the KF/Ca−Mg−Al hydrotalcite catalyst is a very active catalyst with a high active energy in the transesterifications,19 the reaction temperature has a significant influence on the yielding of FAME. In addition, the catalyst is shaped and cut into small cylinders, and the stability of the catalyst is excellent according to our early research.16 As a result, the reduction of FAME yielding rate with high catalyst amount which was reported by Baroutian14 is not found in this case. It also can be seen that the FAME yielding rate increases initially and then decreases with the increase of circulation velocity. This can be explained that the mixing intensity causes the increase of oil conversion when the circulation velocity increases initially, while the reduction in residence time and the decrease in permeate stream when the circulation velocity increases continuously cause the reduction of the FAME yielding amount. 3.3. Optimization. Based on the predictive model, numerical hill-climbing algorithms are used to search for the most desirable outcome. The definition of the optimum conditions is parameters and response with respectively high and low limits to satisfy the creations. And the optimum conditions are listed in Table 3. Experiments are carried out under the optimum conditions in order to evaluate the accuracy of the model. The yielding amount
FAME yielding rate(g/min) = − 0.90393 + 0.00390177 × T − 0.028272 × M + 0.53626 × V + 0.00477401 × T × M − 0.000055 × T × V − 0.013277 × M × V − 0.0000267273 × T 2 − 0.18107 × M2 − 0.083091 × V 2
(2)
The actual values for the yielding rate of FAME in membrane reactor versus the predicted ones according to the developed model equation are shown in Figure 2. The model used to
Figure 2. Yielding rate of FAME predicted from model versus experimental data.
evaluate the FAME yielding rate fits well with the experimental data with a correlation coefficient R2 = 0.9918, which proves that, the interpretation is correct based on the significant effect of factors in the experimental interval and possible interactions. The model can successfully compute the correlations between the process parameters to the yielding rate of FAME. Table 2 presents the analysis of variance (ANOVA) in order to investigate the model fitness and significance. The model F-value Table 2. ANOVA for Response Surface Quadratic Model source
F-value
probability > F
remarks
model T M V TM TV MV T2 M2 V2 lack of fit
390.33 1390.51 666.32 290.54 163.66 0.17 3.16 5.63 25.35 339.99 3.30