Optimisation of integrated biodiesel production. Part II: A study of the material balance

Transcription

1 Bioresource Technology 98 (2007) Optimisation of integrated biodiesel production. Part II: A study of the material balance Gemma Vicente b, *, Mercedes Martínez a, José Aracil a a Department of Chemical Engineering, Faculty of Chemistry, Complutense University, Madrid, Spain b Departament of Chemical and Environmental Technology, Escuela Superior de Ciencias Experimentales y Tecnología (ESCET), Rey Juan Carlos University, Móstoles, Madrid, Spain Received 30 March 2005; received in revised form 4 July 2006; accepted 5 July 2006 Available online 24 August 2006 Abstract A study was made of the material balance for the fatty acid methyl ester (biodiesel) synthesis from sunflower oil using potassium hydroxide as the catalyst. A factorial design of experiments and a central composite design have been used to evaluate the influence of operating conditions on the process material balance. The responses chosen were the biodiesel yield and the yield losses due to triglyceride saponification and methyl ester dissolution in glycerol, while the variables studied were temperature, initial catalyst concentration and the methanol:vegetable oil molar ratio. The biodiesel yield increased and therefore the yield losses decreased by decreasing catalyst concentration and temperature. However, the methanol:sunflower oil molar ratio did not affect the material balance variables significantly. Second-order models were obtained to predict the biodiesel yield and both yield losses. Within the experimental range studied, these models largely matched the results from the experiments. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Biodiesel; Fatty acid methyl esters; Methanolysis; Transesterification; Sunflower oil; Factorial design 1. Introduction The recent rises in petroleum prices and the development of government measures such as The EU Directive 2003/30/EC and The US Energy Policy Act (EPAct) of 1992 have increased the demand for fatty acid methyl esters to be used as fuel in diesel engines and heating systems. In this context, fatty acid methyl esters are referred to as biodiesel. The distinct advantages of biodiesel in comparison with conventional diesel fuel have been shown in part I of this series. Fatty acid methyl esters are synthesized from direct transesterification of vegetable oils, where the corresponding triglycerides react with methanol in the presence of a suitable catalyst. In addition, the process yields glycerol. * Corresponding author. Tel.: ; fax: address: gemma.vicente@urjc.es (G. Vicente). The alkali catalysts are the most commonly used in the biodiesel industry, because the process proves faster and the reaction conditions are moderated (Reid, 1911; Freedman et al., 1984). These catalysts include sodium hydroxide, potassium hydroxide and sodium methoxide. However, sodium methoxide is more expensive than the hydroxides and also more difficult to manipulate since it is very hygroscopic. Potassium hydroxide has the advantage that it can be neutralised with phosphoric acid after the reaction, resulting in potassium phosphate, which may be used as fertiliser. The utilisation of potassium or sodium hydroxide in vegetable oil methanolysis produces soaps by neutralising the free fatty acid in the oil or by triglyceride saponification. Owing to their polarity, the soaps dissolve into the glycerol phase during the separation stage after the reaction. In this sense, both soap formations decrease the biodiesel yield obtained after the separation and purification stages. In addition, the dissolved soaps increase the methyl ester solubility in the glycerol, an extra /$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi: /j.biortech

2 G. Vicente et al. / Bioresource Technology 98 (2007) cause of yield loss (Vicente et al., 2004). The free fatty acid neutralisation can be avoided by using refined vegetable oils (Wright et al., 1944; Bradshaw and Meuly, 1944; Feuge and Gros, 1949; Freedman et al., 1984). However, the most profitable raw materials (e.g. waste cooking oils and fats or low-value fats) usually have a high content of free fatty acid. The yield and yield losses in the biodiesel production process, expressed as % molar and based on the initial amount of vegetable oil, can be evaluated through a material balance of the process. However, there have been a few attempts to study them (Fröhlich and Rice, 1995; Fröhlich et al., 2000; Vicente et al., 2004, 2006). In the present work, the methanolysis of refined sunflower oil was investigated using potassium hydroxide as the catalyst. In this case, therefore, the free fatty acid neutralisation is insignificant. The study is focused on the material balance in order to evaluate the influence of operating conditions on yield and yield losses in the biodiesel production. For this purpose, the factorial design and response surface methodology was utilised. The merits of this methodology are described in part I of this series. In the present study, this technique has been applied to obtain the relationships between the biodiesel molar yield and the molar yield losses (triglyceride saponification and methyl ester dissolution in glycerol) and the operating conditions affecting the transesterification process (temperature, initial catalyst concentration and the methanol:sunflower oil molar ratio). In addition, the methodology was used to calculate the optimum values for these operating variables to obtain the maximum biodiesel molar yield and therefore the minimum values for the molar yield losses. In first part of this study, the factorial design and response surface methodology was applied to obtain the relationship between the biodiesel purity (methyl ester concentration, expressed by % wt) and the operating conditions affecting the transesterification process (temperature, initial catalyst concentration and the methanol:sunflower oil molar ratio) as well as the relationship between the biodiesel weight yield, relative to the initial amount of vegetable oil, and the same operating conditions. The optimum values for these operating variables to obtain the maximum biodiesel purity and yield were a temperature of 25 C, 1.3% wt of potassium hydroxide relative to the amount of vegetable oil and a 6:1 methanol to sunflower oil molar ratio. 2. Methods 2.1. Materials Refined sunflower oil was supplied by Olcesa (Cuenca, Spain). The characteristics of this vegetable oil are described in part 1 of this series, according to the AOCS official methods (1998). Certified methanol of 99.8% purity was obtained from Aroca (Madrid, Spain). The catalyst (potassium hydroxide) was pure grade from Merck (Barcelona, Spain) Equipment Experiments were conducted in a 250 cm 3 three-necked batch reactor, where the total volume of reactives was 125 cm 3. The reactor was equipped with a reflux condenser, a mechanical stirrer and a stopper to remove samples. The reactor was immersed in a constant-temperature bath, which was capable of controlling the reaction temperature to within ±0.1 C of the set point Experimental procedure The reactor was initially filled with the desired amount of oil, then placed in the constant-temperature bath with its associated equipment and heated to a predetermined temperature. The catalyst was dissolved in the methanol and the resulting solution was added to the agitated reactor. The reaction was timed as soon as the catalyst/methanol solution was added to the reactor and it continued for 1 h. The mixture was transferred to a separatory funnel, allowing glycerol to separate by gravity for 2 h. After removing the glycerol layer, the methyl ester layer was washed with two volumes of water to remove methanol and the catalyst and glycerol residuals Analytical methods To calculate the material balance of the reactions, the vegetable oil and the methyl ester and glycerol layers were analysed. The material balance, which refers to the initial amount of vegetable oil, includes the molar yield of biodiesel and the molar yield losses due to triglyceride saponification and methyl ester dissolution in the glycerol phase. Hence, saponification and acid values were determined for the vegetable oils and the methyl esters, according to Cd 3b-76 and Ca 5a-40 AOCS official methods (1998), respectively. Furthermore, 10 g of the glycerol phase was diluted with 30 ml of water and acidified to a ph of less than 2 with 3 M sulphuric acid. The mixture was extracted twice with 20 ml of hexane and once with 20 ml of diethyl ether. The solvents were removed in a rotary evaporator and the residue was dissolved in 50 ml of ethanol. Half of the solution was used to determine its acid value and the remaining half was used to calculate its saponification value (Fröhlich et al., 2000) Statistical analysis The synthesis of biodiesel by sunflower oil methanolysis using potassium hydroxide as the catalyst was developed and optimised following the Factorial Design and Response Surface Methodology. A factorial design was performed to study the effect of the variables on the process and the interaction among variables. The response

3 1756 G. Vicente et al. / Bioresource Technology 98 (2007) Table 1 Levels of factors for the central composite design Factor Levels (+a) (+1) (0) ( 1) ( a) Temperature (T) ( C) Catalyst concentration (C) (%) Methanol:vegetable oil molar ratio (MR) Note: +a: high start point level; +1: high factorial point level; 0: central point level; 1: low factorial point level; a: low star point level. surface methodology was applied to optimise the process. The general flowchart applied in the development of a central composite design is represented in part I of this series. The experimental design in this study was a full 2 3 factorial design (three factors each at two levels). Application of this method requires the adequate selection of responses, factors and levels. The responses selected were the molar yield of biodiesel (MY) and the molar yield losses due to triglyceride saponification (TS) and methyl ester dissolution in the glycerol phase (MEG). All of them are relative to the initial amount of sunflower oil. Selection of the factors and levels was discussed in part I. The levels of the chosen factors are given in Table 1, including the centre and star points. The centre points together with the factorial points make it possible to evaluate the curvature effect. The star points are additional experiments to the factorial and centre points that allow us to determine the nonlinear model. 3. Results 3.1. Linear stage A linear stage was considered in a first step. Table 2 shows the experiments corresponding to the 2 3 factorial design (Experiments 1 8) and four experiments in the centre point to evaluate the experimental error (Experiments 9 12). The first three columns of data give the ±1 and 0 coded factor levels in the dimensionless coordinate, and the following three give the factor levels on a natural scale. Experiments were run at random. Table 2 also shows in the last three columns the results of biodiesel yield and the yield losses due to triglyceride saponification and methyl ester dissolution in glycerol. All of them are expressed as % molar and are relative to the initial amount of sunflower oil. A statistical analysis was carried out on these experimental values, and the main effects and interaction effects of the variables were calculated. The analysis of the main effects and interaction for the three chosen responses together with the test of statistical significance are given in Table 3. The appropriate test is a two-sided t-test with a confidence level of 95%. The temperature and catalyst concentration as well as their corresponding interaction were significant for the three responses analysed because their main effect and interaction values were higher than the corresponding confidence intervals. The methanol:sunflower oil factor was not significant in any of the responses studied since its main effect was lower than the corresponding confidence interval. The other two interactions were less significant or insignificant in every case. Table 2 Experiment matrix and experiment results Stage/type of experiment Run number T ( C) C (% wt) MR X T X C X MR MY (% molar) TG (% molar) MEG (% molar) Nonlinear stage Linear stage Centre points Nonlinear stage Star points a a a a a a Note: T: temperature; C: catalyst concentration; MR: methanol to vegetable oil molar ratio; X: coded value; MY: biodiesel yield; TG: triglyceride saponification; MEG: methyl ester in glycerol.

4 G. Vicente et al. / Bioresource Technology 98 (2007) Table factorial design: statistical analysis (number of experiments: 4; degrees of freedom: 3) Responses Biodiesel yield (% molar) Triglyceride saponification (% molar) Methyl ester in glycerol (% molar) Main effects and interactions I T = I T = I T = I C = I C = I C = I MR = I MR = I MR = I TC = I TC = I TC = I T MR = I T MR = I T MR = I C MR = I C MR = I C MR = Significance test: confidence level: 95% Mean responses Standard deviation t = s = t = s = t = s = Confidence interval ±0.805 ± ± Significant variables T, C, TC, T MR, C MR T, C, TC, C MR T, C, RM,TC, C MR Significance of curvatura Curvature Confidence curvature interval ±0.98 ±0.077 ±0.075 Significance Yes Yes Yes Note: T: temperature; C: catalyst concentration; MR: methanol to vegetable oil molar ratio; I: Main effect or interaction; s: standard deviation. The experimental results were fitted to a linear model, and the following equations were obtained: Statistical models: MY ¼ 95:9350 1:7150X T 2:3825X C þ 0:0475X MR 1:9075X T X C 0:5325X T X RM 0:7350X C X MR ðr 2 ¼ 0:985Þ TS ¼ 1:8600 þ 1:4275X T þ 1:4250X C þ 0:0400X MR þ 1:0775X T X C þ 0:0275X T X MR þ 0:0900X C X MR ðr 2 ¼ 0:998Þ MEG ¼ 0:8575 þ 0:6450X T þ 0:7725X C þ 0:0500X MR þ 0:6000X T X C 0:0125X T X MR þ 0:0800X C X MR ðr 2 ¼ 0:999Þ Industrial models: MY ¼ 85:1125 þ 0:2115T þ 9:6987C þ 1:8104MR 0:1907TC 0:0177T MR 0:9800CMR ðr 2 ¼ 0:985Þ ð4þ TS ¼ 1:4544 0:04187T 2:7187C 0:1346MR þ 0:1077TC þ 0:0009T MR þ 0:1200CMR ðr 2 ¼ 0:998Þ ð5þ MEG ¼ 0: :0252T 1:7950C 0:0546MR þ 0:0600TC 0:0004T MR þ 0:1066CMR ðr 2 ¼ 0:999Þ ð6þ The statistical models were obtained from the coded factor levels and the industrial models from the real values of the factor levels. These equations are only valid within the experimental range studied. The statistical analysis of experiment results also suggests that there is a significant curvature effect for the three chosen responses. Therefore, the first-order polynomial expressions obtained from the first statistical analysis ð1þ ð2þ ð3þ (Eqs. (1) (6)) do not adequately describe the experimental field considered here. It is therefore necessary to consider a more complex design to fit the data to a second-order model in three variables Nonlinear stage As significant curvature effect was detected, four additional runs called star points and coded ±a were added to the 2 3 factorial design plus centre points to form a central composite design. The distance of the star points from the centre point is given by a =2 n/4 (for three factors, a = ). The matrix corresponding to the central composite design is also shown in Table 2, together with the experiment results. The parameters of the second-order model were determined by multiple regression. By considering the coded levels and the real factor levels, expressions were obtained for the statistical model and the technological model, respectively: Statistical models: MY ¼ 98:078 1:4693X T 1:9958X C þ 0:3152X MR 0:9958X 2 T 2:0575X T X C 0:6825X T X MR 0:9888X 2 C 0:5850X CX MR 0:2923X 2 MR ðr 2 ¼ 0:952Þ TS ¼ 0:7577 þ 1:4125X T þ 1:3224X C þ 0:02712X MR þ 0:6220X 2 T þ 1:0775X T X C þ 0:0275X T X MR þ 0:4806X 2 C þ 0:0900X CX MR þ 0:0015X 2 MR ðr 2 ¼ 0:998Þ MEG ¼ 0:1614 þ 0:5293X T þ 0:6508X C 0:0003X MR þ 0:2242X 2 T þ 0:6000X T X C 0:0125X T X MR þ 0:2808X 2 C þ 0:0800X CX MR þ 0:1199X 2 MR ðr 2 ¼ 0:975Þ ð7þ ð8þ ð9þ

5 1758 G. Vicente et al. / Bioresource Technology 98 (2007) Industrial models: MY ¼ 70:3688 þ 0:4928T þ 17:8492C þ 3:5727MR 0:0025T 2 0:2057TC 0:0227T MR 3:9551C 2 0:7800CMR 0:1299MR 2 ðr 2 ¼ 0:952Þ ð10þ TS ¼ 5:7392 0:1826T 6:7690C 0:1515RM þ 0:0015T 2 þ 0:1077TC þ 0:0009T MR þ 1:9225C 2 þ 0:1200CMR þ 0:0007MR 2 ðr 2 ¼ 0:995Þ ð11þ MEG ¼ 5:0749 0:0815T 4:2849C 0:7278MR þ 0:0006T 2 þ 0:0600TC 0:0004T MR þ 1:1232C 2 þ 0:1067CMR þ 0:0533MR 2 ðr 2 ¼ 0:951Þ ð12þ Eqs. (7) (12) describe only the influence of temperature, catalyst concentration and methanol:sunflower oil molar Fig. 2. Response surface plot and contour plot of triglyceride saponification yield loss as function of temperature and catalyst concentration. Methanol:vegetable oil molar ratio of 6:1. Fig. 1. Response surface plot and contour plot of biodiesel yield as function of temperature and catalyst concentration. Methanol:vegetable oil molar ratio of 6:1. ratio on the material balance variables within the studied experimental ranges. For each response, the second-order models can be plotted as three response surfaces graphs and another three contour graphs representing the response (biodiesel molar yield, triglyceride saponification molar yield loss or methyl ester solution in glycerol molar yield loss) as a function of two of the three factors at the centre point value of the third operating condition. Thus, Fig. 1 shows the response surface and contour plots for the predicted values of the biodiesel molar yield as a function of temperature and catalyst concentration at a 6:1 methanol to sunflower oil molar ratio. Fig. 2 illustrates the response surface and contour graphs of the triglyceride saponification molar yield loss predicted for the experimental range of temperature and catalyst concentration at a methanol:sunflower oil molar ratio of 6:1. Finally, the same corresponding response surface and contour plots for the yield loss due to methyl ester dissolution into glycerol is represented in Fig. 3.

6 G. Vicente et al. / Bioresource Technology 98 (2007) due to their polarity. This, in turn, means an increase in the methyl ester solubility in glycerol. As a result, the biodiesel yield decreases Influence of temperature The second factor in importance is the temperature, having a negative influence on the biodiesel yield and a positive one on the biodiesel yield losses. The effect of temperature on the material balance responses is the same as for the catalyst concentration. The biodiesel molar yield also decreases with the temperature due to the increase of triglyceride saponification and the subsequent dissolution of methyl ester into glycerol. In the same way, both yield losses follow a similar tendency since the increase in triglyceride saponification means an increase in the methyl ester solubility in glycerol Influence of the methanol:vegetable oil molar ratio The methanol:vegetable oil molar ratio has a positive influence on the biodiesel molar yield response and a negative one on the biodiesel yield losses, but these influences are insignificant. For this reason, an increase in this variable does not modify these responses significantly Influence of the interactions Fig. 3. Response surface plot and contour plot of methyl ester dissolution in glycerol yield loss as function of temperature and catalyst concentration. Methanol:vegetable oil molar ratio of 6:1. 4. Discussion The effect on the process material balance of the variables, reaction temperature, initial catalyst concentration and methanol:vegetable oil molar ratio is now discussed. The influence of the main factors and interactions are analysed from the statistical models (Eqs. (7) (9)) Influence of catalyst concentration The statistical models show that for the experimental range, catalyst concentration is the most important factor in the biodiesel process. It has a negative influence on the biodiesel molar yield. However, it has an obviously positive influence on the yield losses due to triglyceride saponification and methyl ester dissolution in glycerol. Both yield losses follow a similar negative tendency. In this sense, an increase in the amount of potassium hydroxide increases the amount of soaps produced through triglyceride saponification. The potassium soaps dissolve in the glycerol layer The effect of the temperature catalyst concentration interaction on the material balance responses is very significant. Its influence on biodiesel molar yield is negative, whereas it is positive for the biodiesel yield loss responses. The temperature molar ratio and the catalyst concentration molar ratio interactions have small influences in comparison with the simple effect of the temperature and catalyst concentration and their corresponding interaction Analysis of responses The shapes of the three-dimensional surfaces and contour plots are shown in Figs These graphs represent the material balance responses the biodiesel molar yield, the yield loss for triglyceride saponification and the yield loss for the methyl ester dissolution in glycerol, respectively versus temperature and catalyst concentration at a 6:1 methanol to vegetable oil molar ratio. The graphs in Figs. 2 and 3 have the same shape, since both yield losses followed the same tendency. However, the graphs in Fig. 1 have just the opposite shape because the biodiesel yield followed the opposite trend to the yield losses. The maximum biodiesel molar yield is obtained at low temperatures, low catalyst concentrations and medium methanol:sunflower oil molar ratio. In these operating conditions, the yield losses for the triglyceride saponification and the methyl ester dissolution in glycerol yield losses achieved their minimum values (Figs. 2 and 3, respectively).

7 1760 G. Vicente et al. / Bioresource Technology 98 (2007) material balance responses over the experimental range studied Quality control of biodiesel Fig. 4. Experimental values versus predicted values for the second-order model. Table 4 Quality control of biodiesel from sunflower oil Property Value EU standard EN Viscosity at 40 C (mm 2 /g) 4.1 ± Density at 15 C (g/cm 3 ) 0.88 ± Water content (mg/kg) 270 ± 5 Max. 500 Ester content (% wt) 99.5 ± 0.2 Max MG content (% wt) ± Max. 0.8 DG content (% wt) n.d Max. 0.2 TG content (% wt) n.d Max. 0.2 Free glycerol (% wt) ± Max Total glycerol (% wt) ± Max Acid value (mg KOH/g) ± Max. 0.5 Iodine value (mg I 2 /g) ± 0.7 Max. 120 Note: n.d = not detectable. The plot of experimental values versus predicted is shown in Fig. 4. There are no tendencies in the linear regression fit, so the models adequately described the Some of the most important quality parameters of biodiesel (density, viscosity, water content, ester content, mono-, di- and triglyceride contents, free and total glycerol levels, acid value and iodine value) for reactions using sunflower oil at the selected optimum conditions (25 C, 1.3% wt of potassium hydroxide and 6:1 methanol to sunflower oil molar ratio) are shown in Table 4. These parameters were compared with the following biodiesel standards: European Union Standard EN The table shows the averages of four experiments and the corresponding standard deviations. The density, viscosity and water content of the methyl esters obtained from sunflower oil were within the specifications, indicating that the sunflower oil methanolysis reaction was completed and that the corresponding methyl esters obtained were adequately separated and purified. The ester content was far higher than the only specified limit in the European Union specifications. The levels of individual glycerides (mono-, di- and triglycerides) were also within the specifications, implying that the transesterification reaction was completed. Regarding the free glycerol content, the value measured was lower than its parameter limit and this indicated that the glycerol residuals were eliminated during the purification treatment. Given that the individual glyceride and free glycerol levels were within the specifications, the total glycerol content also met the European standards. The acid values were within specifications in all reactions. The iodine value measured was slightly higher than the specified limit in the European standards (120 mg I 2 /g), because of the high proportion of unsaturated chains in the sunflower oil. 5. Conclusions The main conclusions from this study are the following: A fully central composite design has been applied to evaluate the material balance of the production of fatty acid methyl ester from sunflower oil using potassium hydroxide as the catalyst. In this sense, a study was made of the biodiesel molar yield and the corresponding yield losses (triglyceride saponification and methyl ester dissolution in glycerol). A three-factorial design proved effective in studying the influence of the temperature, initial catalyst concentration and the methanol:vegetable oil molar ratio on the material balance of the process. A response equation has been obtained for the biodiesel molar yield, the saponfication yield loss and the methyl dissolution in glycerol yield loss. From these equations, it is possible to accurately predict the operating conditions required to obtain a given value of the material balance variables.

8 G. Vicente et al. / Bioresource Technology 98 (2007) The investigation of the variables affecting the material balance responses indicated that within the experimental range considered, the most important factor is initial catalyst concentration. This factor has a negative influence on biodiesel molar yield and consequently a positive one on the yield losses for the triglyceride saponification and methyl ester dissolution in glycerol. The temperature is also significant, affecting the responses studied in the same way as initial catalyst concentration. The methanol:sunflower oil molar ratio has no effect on the yield and yield losses. The temperature catalyst concentration interaction also has a negative significant influence on biodiesel yield, but a positive one on both yield losses. A first-order approach did not adequately fit the data for any of the three responses studied and quadratic models were required. Second-order models were developed to predict the biodiesel yield and the yield losses due to triglyceride saponification and methyl ester dissolution in glycerol as a function of the variables. Analysis of the observed values versus the predicted ones demonstrated the efficiency of the models obtained. Acknowledgement This work has been funded by the Comisión Inteministerial de Ciencia y Tecnología from Spain (Projects CI- CYT QUI and CICYT PPQ ). References AOCS, Method Ca 5a-40: free fatty acids. Method Cd 3b-76: saponification value. Method Cd 1-25: iodine value of fats and oils. Wijs method. Method Cd 8-5: peroxide value. In: Firestone, D. (Ed.), Official Methods and Recommended Practices of the American Oil Chemists Society, fifth ed. American Oil Chemists Society, Champaign, IL, USA. Bradshaw, G.B., Meuly, W.C., Preparation of detergents. US Patent 2,360,844. Feuge, R.O., Gros, A.T., Modification of vegetables oils. VII. Alkali catalyzed interesterification of peanut oil with ethanol. J. Am. Oil Chem. Soc. 26 (3), Freedman, B., Pryde, E.H., Mounts, T.L., Variables affecting the yields of fatty esters from transesterified vegetable oils. J. Am. Oil Chem. Soc. 61 (10), Fröhlich, A., Rice, B., The preparation and properties of biodiesel grade methyl ester from waste cooking oil. In: Minutes of the Activity Meeting of the IEA, Vienna, pp Fröhlich, A., Rice, B., Vicente, G., The conversion of waste tallow into biodiesel grade methyl ester. In: 1st World Conference and Exhibition on Biomass for Energy and Industry. Proceedings of the Conference held in Seville (Spain), pp Reid, E.E., Studies in esterification. IV. The interdependence of limits as exemplified in the transformation of esters. Am. Chem. J. 45, Vicente, G., Martínez, M., Aracil, J., Integrated biodiesel production: a comparison of different homogeneous catalysts systems. Bioresour. Technol. 92 (3), Vicente, G., Martínez, M., Aracil, J., A comparative study of vegetable oils for biodiesel production in Spain. Energy Fuels 20, Wright, H.J., Segur, J.B., Clark, H.V., Coburn, S.K., Langdon, E.E., DuPuis, R.N., A report on ester interchange. Oil Soap 21,