Modeling and Optimization of Catalytic Dehydration of Ethanol to Ethylene Using Central Composite De

Trans. Tianjin Univ. 2009, 15: 366-370DOl10.1007/s1220900900648Modeling and optimization of catalyticDehydration of Ethanol to Ethylene UsingCentral Composite DesignKONG Haining(孔海宁), QI Ers(齐二石)1, LI Gang(李钢)',HE Shuguang(何曙光)1, ZHANG Xian(张宪)2(1. School of Management, Tianjin University, Tianjin 300072, China;2. School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China)Abstract: The central composite design in the modeling and optimization of catalytic dehydration of ethanol toethylene was performed to improve the ethylene yield. A total of 20 experiments at random were conducted to in-vestigate the effect of reaction temperature, Si/Al ratios of H-ZSM-5 catalyst and liquid hourly space velocity(LHSV) on the ethylene yield. The results show that the relationship between ethylene yield and the three signifi-cant independent variables can be approximated by a nonlinear polynomial model, with R-squared of 99.9%andadjusted R-squared of 99.8%. The maximal response for ethylene yield is 93. 4% under the optimal condition of 328C. Si/Al ratio 85 and LHSV 3.8h 1Keywords: central composite design: catalytic dehydration of ethanol; ethylene yield; modeling; optimizationEthylene is commercially produced by the thermal LHSv on the ethylene yieldcracking of liquefied petroleum gas(LPG) or naphthaThe process is an endothermic reaction requiring high 1 Experimenttemperature of 600-1 000 C. Compared with the con-ventional route, the catalytic dehydration of ethanol to 1.1 Materialsethylene is attractive since it requires lower temperature Ethanol absolute was provided by Tianjin Bench-and offers higher ethylene yield. Moreover, ethanol can mark Chemical Reagent Co, Ltd, and a series of H-ZSM-be produced from renewable sources, independent of the 5 zeolite with different Si/Al ratios was provided bypetroleum source. The dehydration of ethanol to ethylene Catalyst Plant of Nankai University. All solvents andcan be catalyzed using different catalysts, such as x other chemicals were of analytical grade and provided byalumina, titanium oxides), magnesium oxides+, cobalt Tianjin Fuchen Chemical Reagent Co, Ltdoxides'5)and H-ZSM-5 zeolite(a type of zeolites)(6). Ear- 1.2 apparatuslier researches revealed that H-ZSM-5 catalyst is an effi-The experimental setup is shown in Fig. 1. The catacient catalyst and the reaction temperature, Si/Al ratios lytic reaction studies were performed in electricallyand liquid hourly space velocity(LHSV) have a great heated fixed bed reactor(inside diameter 10 mm,lengtheffect on the ethylene yield! -). The experimental design 420 mm, from Tianjin Zhonghuan Instrument Co, Ltd)is efficient to estimate the effects of several variables Ethanol was poured through a micro pump. Usually theRecently, it has been successfully reactor contained 3 mL catalyst. The reaction temperatureapplied to a different process for achieving its optimiza- was varied in the range of 216-314 C. The range oftion. The main purpose of this paper is to perform si/Al ratios of catalysts was 33-117. LHSV values werethe central composite design in order to investigate the in the range of 1.5-65h.effect of Si/Al ratios, reaction temperature, and中国煤化工Accepted date: 2009-03-19CNMHGSupported by National Natural Science Foundation of China(No. 70671UKONG Haining, bom in 1981, female doctorate student.CorrespondencetoQIErshi,E-mail:qies@tju.edu.cnKONG Haining et al: Catalytic Dehydration of Ethanol to EthylenGas chromatograph1.3 Analysis of productsb2x x2+i x,X,+b2x2x,The products were analysed on gas chromatograph where Y is the predicted response(ethylene yield); bo isSP-3420( Bejing Beifen Instrument Co, Ltd )with flame the average of the results of all experiments; b, bz and byionization detector( FID)using PEG 20 000 capillary are the main half-effects of the coded factors X, X, andcolumn by programming the oven temperature in the X, respectively; bu, bz and ba3 are the quadratic effectsrange of60-10℃atl℃/minb12, b13 and b23 are two-factor interaction half-effects1.4 Experimental designThe design and analysis of the central compositeThe central composite design is an ettective experiment were carried out with Minitab. The experi-higher-order experimental design method, which seeks ments were conducted in a randomized order to avoidthe optimal conditions for a multivariable system, butsystematic bias. The central composite experimental de-requires much fewer experiments than a full factorialsign is shown in Tab. 2design method and has been shown to be sufficient todescribe the majority of steady-state process respTab2 Central composite experimental designThis paper performs the central composite design of three Standard Runorderorderfactors. The 20 experiments were performed at randorThe central composite experimental factors and level val68272.6ues are shown in Tab. 15341.6820Tab1 Experimental factors and levels in dehydration ofethanol to ethylene68.915913Factor85.72162503003501.68275.720The factors in this experiment are X (reaction tem-0010001010001000perature, C), X2(Si/Al ratios)and X,(LHSV,hThe central composite design is mainly employed in or-16-1.682der to produce a detailed model(Eq (1)), and investigateits linear, interaction and higher-order(quadratic) terms中国煤化工-1682774Y=bo+bx,+b2x2+b,x,+CNMHG88.7b1X2+b2X2+b3x2+Transactions of Tian in UniversityNo.520092 Results and discussionR-sq=Using Minitab 14.0 statistical software, the experi-mental results are fitted to a second-order model. Thecoefficients of the model and p value are shown in Tab. 3With a 95% confidence, blocks, XiX3 and X2x3 are insig-nificant. The model is as follows:Y=90.7+10.8X1-23X2Fig 2 Relationship between experimental and predicted15X3-116X2-09X2-54X3+54Xx1x2Tab3 Estimated regression coefficients for ethylene yieldCoefficientf-testConstan90.73330.276528.1100.0001083130.1835590350.1835-12.75-14550.18350.1786Fig 3 Normal probability plot of residuals for F0.1786-5.1450.00(ethylene yield)in Eq- (2)0.1786XIX,5.40000.2397225260.0000.12500.00023972.1 Validity of modelFig 2 shows the correlation between experimentaland predicted values using Eq (2). The coefficient ofvariation R-squared(R-Sq)and adjusted R-squared( R-SqFitted value/%(ad))are 99.9%and.8% respectively, indicating a highdegree of correlation between experimental and predictedFig 4 Residuals vs. fitted values for Y(ethylene yield)in Eq- (2)valuesFig3 is the normal probability plot. It shows that the 2.2 Analysis of variancrequency distribution of residuals for ethylene yield ap-The analysis of variance for response r(ethyleneproximately follows the fited normal distribution curve. yield )is shown in Tab. 4. Tab. 4 shows that the linear ef-Fig 4 is the residuals vs. the fitted values for Y(ethylene fects the quadratic effects and the interactions betweenyield)in Eq- ( 2). As can be seen, the residuals of y(ethyl- the factors are significant (P<0.05), but the lack-of-fitene yield) randomly scatter in the residual plotsinsignificant(P>0.05)Tab 4 Analysis of variance for ethylene yieldAdjusted sum ofdjusted mean of4140.041000610.0001705891705892200.752200.75733.5821595.70000016924Lack-of-fit30553.72中国煤化工42Pure error4144.64YCNMHG-368KONG Haining ef al: Catalytic Dehydration of Ethanol to Ethylene2.3 Response surface plots and optimization condi- X3 (0.14), i.e. the ethylene yield is 93.4% under the con-dition of reaction temperature 328C, SiAl ratio 85, andThe response surface plot for the predicted response LHSV 3. 8 h. Three experiments were conducted atY (ethylene yield), based on the second-order model is these optimal variables. The average response is 93.5%,shown in Fig. 5. Fig. 5 shows the relationship between two which is well fitted with the result of response optimizavariables and response y(ethylene yield)at the centre tionlevel of other variables. As can be seen in Fig. 5, the rela-tionship between response y(ethylene yield) and vari- 3 Conclusionsables xx, xa is nonlinearAs shown in Fig. 5, there are maximum points for r The catalytic dehydration of ethanol to ethylene isresponse in the response surface plots. Therefore, the studied in this paper. The mathematical model is estab-response surface optimization could be found when the lished for ethylene yield by using sets of experimentartial derivative of y with respective to x, is equal to 0. data and Minitab. The predicted values match the exith R-Sq of 99. 9% and R-Sq(adj)of 99.8% for response Y. The result of optimization is thatthe maximal ethylene yield is 93.4% under the optimalcondition of 328C Si/Al ratio 85 and LHSV.8b.References[1]20fins by Catalytic Partial Oxidation of Hydrocarbons[P]US3541179.1970-11-17[2]Winters O, Eng M. Make ethylene from ethanol [J]. Hy(a)y vs. xi and xdrocarbon Processing, 1976, 55(11): 125-133[3]Gao X T, Wachs I E. Titania-silica as catalysts: Molecular[]. Catalysis Today, 1999, 51: 233-254[4 Golay S, Kiwi-Minsker L, Dopper R et al. Influence ofthe catalyst acid/base properties on the catalytic ethanolChemical Engineering Science, 1999, 54: 3593-3598.[5]Doheim MM, EI-Shobaky H G. Catalytic conversion ofethanol and iso-propanol over Zno-treated CoyOwALO3(b)y vs, xI and x1solids []. Colloids and Surfaces A: Physicochemical andEngineering Aspect, 2002, 204: 169-174.[6 Le Van Mao R, Dao L H Ethylene Light Olefins fromEthanol[P].Us4698452.1987-10-06[7] Nguyen T M, Le Van Mao R. Conversion of ethanol inaqueous solution over ZSM-5 zeolites: Study of the reac-tion network [J]. Applied Catalysis, 1990, 58: 119-129[8 Schulz J, Bandermann F Conversion of ethanol over zeolite H-ZSM-5 [J]. Chemical Engineering and Technology,Fig 5 Surface plots994,17:179-186.[9] Phillips C B, Datta R. Production of ethylene from hydrousThe result of the response optimization of nonlinear中国煤化工 ditions [j. industrialpolynomial model is as follows. The response Y reachesCNMHGrch 1997, 36(11)maximal (93. 4%)with parameters of X1(0.56), X2(0.39),Transactions of Tianjin University VoL 15 No 5 200910] Box GEP, Hunter W G, Hunter JS. Statistics for Experi- 14] Tsapatsaris S, Kotzekidou P. Application of central commenters:An Introduction to Design, Data Analysis andposite design and response surface methodology to theModel Building [M]. John wiley, New York, 1978fermentation of olive juice by Lactobacillus plantarum and11] Deming S N, Palasota J A, Palasota J M. ExperimentDebaryomyces hansenii [J]. International Journal of Fooddesign in chemometrics []. Journal of Chemometrics,Microbiology, 2004, 95: 157-1681991,5:181-192.[15] BoxGEP, Wilson K G On the experimental attainment12] Capella-Peiro M E, Bose D, Rubert M F et al. Optimiza-tion of a capillary zone electrophoresis method by using aSociety, Series B: Statistical Methodology, 1951, 13: 1-4central composite factorial design for the determination of [16] Box G E P, Hunter JS. Multifactor experimental designscodeine and paracetamol in pharmaceuticals [J].Journalfor exploring response surfaces [J]. Annals of Mathemati-of chromatography B, 2006, 839: 95-101cal Statistics, 1957. 28: 195-24113]Avila A, Sanchez E I, Gutierrez M I. Optimal experimental [17] NIST. SEMATECH E-Handbook of Statistical Methodsdesign applied to the dehydrochlorination of poly (vinylLeb/ol].http://www.itlnistgov/div898/handbook/pri/loride)[J]. Chemometrics and Intelligent Laboratorysection3/pri3361. htm. 2006-06-18Systems,2005,77:247-250.中国煤化工CNMHG370