Diponegoro University From the SelectedWorks of Istadi May 12, 2008 Optimization of Methane Conversion to Liquid Fuels over W-Cu/ZSM5 Catalysts by Response Surface Methodology Istadi Istadi, Diponegoro University Available at: https://works.bepress.com/istadi/14/
Journal of Natural Gas Chemistry 17(2008)39 44 Optimization of methane conversion to liquid fuels over W-Cu/ZSM-5 catalysts by response surface methodology Didi Dwi Anggoro, Istadi Chemical Reaction Engineering and Catalysis Group, Department of Chemical Engineering, University of Diponegoro, Tembalang, Semarang 50239, Indonesia [ Manuscript received August 30, 2007; revised January 7, 2008 ] Abstract: The conversion of methane to liquid fuels is still in the development process. The modified HZSM-5 by loading with Tungsten (W) enhanced its heat resistant performance, and the high reaction temperature (800 C) did not lead to the loss of W component by sublimation. The loading of ZSM-5 with Tungsten and Copper (Cu) resulted in an increment in the methane conversion, CO 2,andC 5+ selectivities. The high methane conversion and C 5+ selectivity, and low H 2 O selectivity are obtained by using W/3.0Cu/ZSM-5. The optimization of methane conversion over 3.0 W/3.0Cu/ZSM-5 under different temperature and oxygen concentration using response surface methodology (RSM) are studied. The optimum point for methane conversion is 19% when temperature is 753 C, and oxygen concentration is 12%. The highest C 5+ selectivity is 27% when temperature is 751 C, and oxygen concentration is 11%. Key words: methane; W-Cu/ZSM-5; liquid hydrocarbons; response surface methodology 1. Introduction Corresponding author. Tel: 6224-7460058; Fax: 6224-7460055; E-mail: anggoro@alumni.undip.ac.id The use of ZSM-5 zeolite as a support of the metal oxide phase is very interesting due to three reasons: its thermal stability, the high surface area that enables high metal oxide loading, and the presence of acid sites that could lead to the formation of certain active metal oxide species [1]. Cu loaded ZSM-5 catalyst via acidic ion exchange method has been identified to be the potential catalyst for conversion of methane to liquid fuels [2]. However, the infrared study of metal loaded ZSM-5 catalyst indicated that the catalysts are not resistant to high temperature. Previous studies have indicated that metal loaded ZSM-5 did not exhibit vibration band at 3610 cm 1 and 3660 cm 1, except for ZSM-5 which showed a weak vibration band at 3666 cm 1. The result suggested that the framework and nonframework aluminum were either extracted to acidic solution or became silanol defect form when calcined at 800 C and made the catalysts inactive [3]. Previous studies [4,5] indicated that the Cu loaded W/ZSM-5 catalyst was thermally stable at the reaction temperature (700 800 C). In our previous study [5], it was reported that the loading of HZSM-5 with tungsten and copper decreased the crystallinity, surface area, and also total volume of the catalysts. However, the average pore diameter and the acidity of the zeolites increased as a result of the modification with the metals. The modified ZSM-5 by loading with Tungsten enhanced its heat resistant performance and the high reaction temperature (800 C) did not lead to loss of W component by sublimation. The process of converting methane to liquid hydrocarbons (C 5+ ) is dependent on the metal surface area and the acidity of the zeolite. The W/3.0Cu/ZSM-5 is the potential catalyst, because over this catalyst high methane conversion and C 5+ selectivity, and low H 2 O selectivity are obtained. Amin and Anggoro [5] studied the optimization of Cu loaded W/ZSM-5 using the Response Surface Methodology. The low, middle, and high levels of all the independent variables were W content in weight doped into the 3.0Cu/W/ZSM-5, O 2 concentration, and flow rate of feed gases. This article reports that the optimizations of W loaded Cu/ZSM-5 with the independent variables were temperature and oxygen concentration. Response surface methodology (RSM) is a method to determine the optimum condition of a process. RSM has similarity with regression analysis. In regression analysis, empirical mathematical model are derived from the experiment
40 Didi Dwi Anggoro et al./ Journal of Natural Gas Chemistry Vol. 17 No. 1 2008 data. RSM is a set of techniques designed to find the optimum value of the response and the influencing factors. RSM technique has been successfully applied in the field of quality experimental work [5 12]. This article estimates the optimization of temperature and oxygen concentration on conversion of methane to liquid fuels using response surface methodology (RSM). The temperature and oxygen concentration used were 700, 750, 800 Cand 5%, 10%, 15%, respectively, resulting in a two-level experimental design, with 10 experiments. Empirical parameters were estimated with standard least-square procedure, using the Design Expert Statsoft software Statistica version 6.0 2001. 2. Experimental 2.1. Preparation, characterization, and testing of catalysts ZSM-5 zeolite with a SiO 2 /Al 2 O 3 mole ratio of 30 was supplied by Zeolyst International Co, Ltd, Netherlands. The surface area of the zeolite is 400 m 2 /g. The W (3% weight)-hzsm-5 catalyst was prepared by impregnating a certain amount of the HZSM-5 zeolite carrier with ammonium tungstate hydrate solution. X-ray diffraction (XRD), H 2 -temperature programmed reduction (H 2 -TPR), NH 3 -temperature programmed desorption (NH 3 -TPD), N 2 adsorption, and FT-IR carried out the characterization of catalysts. XRD and FT-IR were used to determine the zeolite structure. NH 3 -TPD provided the acidity of catalyst samples. H 2 -TPR data were pertinent to the the zeolite morphology. The preparation and characterization of catalysts were explained in previous articles [4]. The catalysts were tested for methane conversion to liquid hydrocarbons (LHC) via a single step reaction in a fixedbed micro reactor. Methane with 99.9% purity was reacted at atmospheric pressure and various temperature and oxygen concentration. An on-line Gas Chromatograph with TCD and Porapak-N column was utilized to analyze the gas. The liquid product was analyzed using GC FID and HP-1 column. 2.2. Optimization procedure The empirical models were used to analyze the influence of the process variables on the response factors. Second order polynomial models were used to verify the linear and quadratic effects of the process variables and their linear and quadratic interactions. The study used normalized process variables, in order to compare the relative importance of model parameters and to obtain independent linear parameter. Therefore, process variables assumed the values ( 1), (0), or (+1) for minimum, central, and maximum. For temperature as an example, the values of 700, 750, and 800 C stand for minimum, central, and maximum temperature, respectively (Table 1). The optimization method based on RSM involved three major steps: design of experiment using statistical approach, coefficient estimation based on mathematical model and response prediction, and finally model adequacy check. The equation model is tested with analysis of variance (ANOVA) with 99% degree of confidence. The RSM output such as contour and 3D graphic surface plots provide the optimum and most influential variable for methane conversion and C 5+ selectivity. According to central composite design, the total number of experiment combinations is 2 k +2k + n 0,wherek is the number of independent variables and n 0 is the number of experiments repeated at the center point [13 16]. In this case, n 0 =2. The variables X i were coded as x i according to Equation (1). The basis of forming a polynomial equation is given in Equation (2): Y u = β 0 + x i = X 1 X 0 ; i = 1,2,3,...,k (1) X k i=1 β i X ui + k i=1 β ii X 2 ui + k β ij X ui X uj i<j (2) with Y u predicted response u; β 0 offset term; β i linear term; β ii squared term; β ij interaction term; x i dimensionless value of an independent variable; X i real value of an independent variable; X 0 real value of an independent variable at center point; X step change and u = 1, 2,..., n. The low, middle, and high levels for all these variables (independent variables) were based on prior screening from literature review and accordingly, temperatures of 700, 750, and 800 C were chosen for variable X 1 (operating temperature) and 5%, 10%, and 15% for X 2 (volume percent of oxygen on gas feed) as shown in Table 1. Allowances for extreme measures are designated in central composite design and presented by α and +α, as shown in Table 1. Table 1. The levels of variables chosen for trials Temperature, X 1 ( C) Oxygen, X 2 (%) α 679.3 2.9 1 700.0 5.0 0 750.0 10.0 +1 800.0 15.0 +α 820.7 17.1 The actual design experiment is shown in Table 2. It was found that a total of 10 runs were necessary in order to optimize processing parameters for enhanced methane conversion and C 5+ selectivity. The design experiment was carried out using the Design Expert Statsoft software Statistica version 6.0 2001.
Journal of Natural Gas Chemistry Vol. 17 No. 1 2008 41 Table 2. Experimental plan for optimization process Run no Independent variables Temperature ( C) Oxygen on feed (%) 1 700.0 5.0 2 700.0 15.0 3 800.0 5.0 4 800.0 15.0 5 750.0 10.0 6 679.3 10.0 7 820.7 10.0 8 750.0 2.9 9 750.0 17.1 10 750.0 10.0 3. Results and discussion Figure 1. Observed versus predicted values for CH 4 conversion The results of this study showed that the conversion of methane and liquid hydrocarbon (C 5+ ) selectivity depends on the operating temperature and oxygen concentration. RSM design for two independent variables was taken to obtain the combination of values that optimizes the response within the region of the 3-D observation space, which allows one to design a minimal number of experimental runs. The model also evaluated the effect of each independent variable to a response, singly and in combination with other variables, which is not otherwise feasible. The experimental value and predicted response for the 10 trial runs carried out are presented in Table 3. Table 3. Experimental and theoretical predicted values for CH 4 conversion and C 5+ selectivity Run CH 4 conversion (%) C 5+ selectivity (%) no Experimental Predicted Experimental Predicted 1 14 14.116 10 12.387 2 16 16.280 18 17.362 3 13 13.719 6 8.887 4 16 16.884 20 19.862 5 17 18.500 23 26.500 6 16 15.927 21 20.228 7 17 16.073 21 19.521 8 13 12.616 7 3.736 9 17 16.384 14 15.014 10 20 18.500 30 26.500 The illustration of the observed and predicted values can be referred to Figures 1 and 2. From both Figures, it can be seen that most of the points of experimental values lies close to the straight line which is the predicted values. From Table 3, indicated that the different value of methane conversion by experimental with predicted is around 0.073% 1.500%. The different value of C 5+ selectivity by experimental with predicted is around 0.137% 3.500% The highest deviation for methane conversion occurs at run number 10 where the value is 1.500. The highest deviation for C 5+ selectivity occurs at run number 10 where the value is 3.500. Figure 2. Observed versus predicted values for C 5+ selectivity In the present investigation, it was observed that the responses changed significantly with variation in the three variables. From the experimental data, the highest methane conversion and C 5+ selectivity were 20% and 30%, respectively. Using multiple regression analysis on the experimental data, the following second order polynomial equations were found to explain the value of both dependent variables. In this work, the number of independent variables are two and therefore, k = 2. Equation (2) becomes: Y 1 = 1302.10 + 3.50103X 1 + 2.07X 2 0.00235X 2 1 0.1648X2 2 + 0.030X 1X 2 (3) Y 2 = 2712.77 + 7.0975X 1 + 20.408X 2 0.004675X1 2 0.3776X2 2 + 0.014X (4) 1X 2 where Y 1 is the predicted methane conversion and Y 2 is the predicted C 5+ selectivity. The analysis of variance (ANOVA) Table displaying the total regression, and error of sum of squareisshownintables4and5. The F value (Tables 4 and 5) is a ratio of the mean square due to regression to the mean square due to error. The value of F is compared to the table value F (p 1, N p, α), p is the
42 Didi Dwi Anggoro et al./ Journal of Natural Gas Chemistry Vol. 17 No. 1 2008 Table 4. ANOVA for the CH 4 conversion (quadratic response surface model fitting) Source Sum of Degree of Mean squares freedom square F Value F (6,4,0.01) R 2 S.S. regression 33.618 6 5.603 3.077 7.9761 0.822 S.S. error 7.282 4 1.821 S.S. total 40.900 10 S.S. Sum of squared shows similar plot of C 2 selectivity for oxygen content and temperature. From RSM (Statistic software), the optimum point for methane conversion is 18.725% when temperature is 752.718 C and oxygen concentration is 11.682%. The highest C 5+ selectivity is 26.965% when temperature is 750.764 C and oxygen concentration is 11.171%. Table 5. ANOVA for the selectivity for liquid hydrocarbons (quadratic response surface model fitting) Source Sum of Degree of Mean squares freedom square F Value F (6,4,0.01) R 2 S.S. regression 472.572 6 78.762 5.897 7.9761 0.898 S.S. error 53.428 4 13.357 S.S. total 526.000 10 S.S. Sum of squared number of the terms in the fitted model, and N is a number of run experiment. If the value of F is smaller than F (p 1, N p, α), then the null hypothesis is accepted at the α level of significance. If the null hypothesis is true, it means that the model is a good predictor of the experimental data. From ANOVA table, the value of F for methane conversion and C 5+ selectivity are 3.077 and 5.897, respectively. These F values are smaller than the tabulated F (p 1, N p, α),which is 7.9761. This means that the model for predictor of the experimental data (Equations 3 and 4) is true. The coefficient of determination R-squared (R 2 )(Tables 4 and 5) for methane conversion and C 5+ selectivity are 0.822 and 0.898, respectively. The value of R 2 is a measure of total variation of observed values about the mean explained by the fitted model. The value is always between zero and one. A value of one indicates that the statistical model explains all of the variability in the data. A value of zero indicates that none of the variability in the response can be explained by the experimental factor. Hence, the values of R 2 in this study that is greater than 80% show a good agreement between experimental data and predicted values. The fitted model from Equations (1) and (2) can be used to map empirically the response function over the experimental region. The contour plot helps in assessing the effect of any two variables in combination on the product quality. The 3-D graphical surface plot illustrates the response value of dependent variables. The effects of operating temperature and oxygen concentration (Figure 3) on the methane conversion can be obtained. Similarly, the effects of operating temperature and oxygen concentration (Figure 4) on the C 2 selectivity can be obtained as well. The coordinates of the stationary point (Figures 3 and 4) are called the optimum point. Figure 5 shows the 3-D graphical surface plot of methane conversion for oxygen content and temperature. Figure 6 Figure 3. Contour surface plot of methane conversion as a function of temperature and oxygen concentration Figure 4. Contour surface plot of C 5+ selectivity as a function of temperature and oxygen concentration Figure 5. 3-D graphic surface optimization of methane conversion versus temperature and oxygen concentration
Journal of Natural Gas Chemistry Vol. 17 No. 1 2008 43 Figure 6. 3-D graphic surface optimization of C 5+ selectivity versus temperature and oxygen concentration Figures 7 and 8 demonstrated the Pareto chart analysis result. The normal probabilities of the rank-ordered parameters are plotted on the y-axis, and the actual parameter estimates (optionally standardized) are plotted on the x-axis. Figure 7. Pareto chart of effect of variables for methane conversion. Concn. = concentration, Q: Quadratic, L: Linear The p-level of 0.05 (i.e, 1/20) indicates that there is a 5% probability that the relation between the variables found in our sample is a fluke. In other words, assuming that in the population there was no relation between those variables whatsoever, and we were repeating experiments like ours one after another, we could expect that approximately in every 20 replications of the experiment there would be one in which the relation between the variables in question would be equal or stronger than in ours. The Pareto chart for methane conversion (Figure 7) shows clearly that a percentage of O 2 on feed gas (quadratic) is the most important and influential factor in determining the value of response variable with effect values of 3.15. Likewise, Pareto chart for C 5+ selectivity (Figure 8) shows clearly that oxygen concentration on gas feed (quadratic) is the most vital in determining the value of response variable with effect values of 4.966. 4. Conclusions The Response Surface Methodology involving an experimental design and regression analysis is effective in finding the optimum point of the independent variables, and in assessing their effects on the two responses considered. The methane conversion and C 5+ selectivity are determined by the operating temperature, and oxygen concentration on gas feed. This shows that variable factor and level selection meets experiment designing concept. The second order polynomial equation model estimation whose validity is agreed upon is estimated using ANOVA statistical testing and yields 99% degree of confidence of response behaviors to variables. The experimental design and empirical modeling was possible to optimize the oxidation of methane to liquid hydrocarbons. The optimum point for methane conversion is 19% when temperature is 753 C, and oxygen concentration is 12%. The highest C 5+ selectivity is 27% when temperature is 751 C, and oxygen concentration is 11%. The oxygen concentration on gas feed presented the most important effects on the response factor. The temperature reaction had a limited effect probably due to the reason that the reaction is exothermic reaction. References Figure 8. Pareto chart of effect of variables for C 5+ selectivity. Concn. = concentration, Q: Quadratic, L: Linear [1] de Lucas A, Valverde J L, Canizares P, Rodriguez L. Appl Catal A: General, 1998, 172(1): 165 [2] Amin N A S, Anggoro D D. J Natur Gas Chem, 2003, 12(2): 123 [3] Amin N A S, Anggoro D D. J Natur Gas Chem, 2002, 11(1 2): 79 [4] Anggoro D D, Amin N A S. J Natur Gas Chem, 2006, 15(4): 340 [5] Amin N A S, Anggoro D D. Fuel, 2004, 83(4 5): 487
44 Didi Dwi Anggoro et al./ Journal of Natural Gas Chemistry Vol. 17 No. 1 2008 [6] Kasankala L M, Xue Y, Weilong Y, Hong S D, He Q. Bioresource Technology, 2007, 98(17): 3338 [7] Mundra P, Desai K, Lele S S. Bioresource Technology, 2007, 98(15): 2892 [8] Pio T F, Macedo G A. Enzyme Microb Technol, 2007, 41(5): 613 [9] Körbahti B K, Aktaş N, Tanyolaç A. J Hazardous Materials, 2007, 148(1 2): 83 [10] Macedo G, Pio T. J Biotechnology, 2007, 131(2): S211 [11] Ushakumari S R, Rastogi N K, Malleshi N G. J Food Engng, 2007, 82(1): 35 [12] Psomas S K, Liakopoulou-Kyriakides M, Kyriakidis D A. Biochemical Engineering Journal, 2007, 35(3): 273 [13] Murphy T D. Chem Engng, 1977, 6: 168 [14] Burtis C A, Bostick W D, Overton J B, Mrochek J E. Anal Chem, 1981, 53(8): 1154 [15] Cornell J A. How to Apply Response Surface Methodology. USA: American Society for Quality Control. 1990 [16] Box G E P, Hunter W G, Hunter J S. Statistics for Experiment: an Introduction to Design, Data Analysis and Model Building. New York: Wiley. 1978