Published on: **Mar 3, 2016**

- 1. Kinetics and Fixed-Bed Reactor Modeling of Butane Oxidation to Maleic Anhydride Ramesh K. Sharma and David L. Cresswell Systems Engineering Group, Federal Institute of Technology, CH-8092 Zurich, Switzerland Esmond J. Newson Swiss Aluminium R&D, CH-8212 Neuhausen, Switzerland Selective oxidation kinetics of n-butane to maleic anhydride in air were studied over a commercial, fixed-bed vanadium-phosphor oxide catalyst. The temperature range was 573-653 K with butane concentrations up to 3 mol % in the feed, which is within flammability limits but below ignition temperatures. The rate data were modeled using power law kinetics with product inhibition and included total oxidation and decomposition reactions. Kinetic parameters were es- timated using a multiresponse, nonlinear regression algorithm showing intercorre- lation effects. The kinetics were combined with independent measurements of catalyst diffusivity and reactor heat transfer using a one-dimensional heterogeneous reactor model. Model predictions and observed temperatures and concentrations from non- isothermal pilot plants were compared up to 115 days on stream. Agreement was acceptable with inlet butane concentrations up to 2.7 mol %. For example, runaway was predicted at a salt temperature 3 K higher than observed. Effectiveness factors around the hot spot were estimated at 0.6 with the catalyst surface temperature 2- 3 K higher than the average gas temperature. Introduction For economic and ecological reasons, n-butane is replacing runaway situations in commercial pilot-plant reactors. benzene as the major feedstock for the production of maleic Prior kinetic studies have been undertaken from either of anhydride. Enhanced catalyst activity and selectivity (Budi et two experimental viewpoints. The first and more fundamental al., 1982),together with anhydrous, organic solvent absorption approach is the identification of intermediate surface species downstream (Neri and Sanchioni, 1982), make the economics and their relative concentrations to understand the reaction for the fixed-, or fluidized-bed process exbutane particularly mechanism and different reaction pathways (Centi et al., 1988; favorable. More recently, a recirculating solids reactor has been Hodnett, 1987; Wenig and Schrader, 1986). Many such pub- described, which optimizes the best features of fixed and flui- lications, however, fail to quantify kinetics for subsequent dized beds (Contractor et al., 1988). reaction engineering and reactor optimization. The alternate Whatever the preferred technology, knowledge of the in- approach is to obtain intrinsic reaction data, which is then trinsic kinetics is essential for optimizing the reactor design confronted with different empirical and mechanistic models (Schneider et al., 1987; Wellauer et al., 1986). For safety in (Centi et al., 1985). This approach requires statistical inter- reactor operation with respect to temperature runaway at high pretation of the data to discriminate among different models butane inlet concentrations, a study has been made (Sharma and to show intercorrelation of parameters thus selecting the et al., 1984b), whose kinetics were used for comparison of most probable kinetic parameters (Schneider et al., 1987; Sharma et al., 1984b). The latter approach was chosen for this work whose purpose Correspondence concerning this article should be addressed to E. J. Newson, Paul Scherrer Institute, Dept. FS, CH-5232 Villigen PSI, Switzerland. is to develop an intrinsic kinetic model, which is not only Present address: R. K. Sharma, Dept. of Chemical Engineering, University of Saskatch- consistent with laboratory and pilot-plant kinetic measure- ewan, Saskatoon, Canada, S7N OWO: D. L. Cresswell, Chemicals and Polymers Ltd., P.O. Box 8, The Heath, Runcorn, Cheshire, England, WA7 4QD. ments but also capable of being combined with heat and mass AIChE Journal January 1991 Vol. 37, No. 1 39
- 2. transport steps to predict nonisothermal, fixed bed, pilot-plant CAHIO+ 3.50z-c4H203 i - 4 H z 0 (1) data at commercial conditions. The kinetic work was accom- plished in three stages first using isothermal, differential rate C4HZ03 p 0 2 - ( 6 - 2p)CO + (2p - ;!)CO + + HzO (2) data, then isothermal integral rate data from pilot plants, and finally testing the scale-up properties of the reactor model C4H10+n02-(13-2n)CO+(2n- 91CO2 L ~ H , O (3) against full-scale pilot-plant data. The major results of this work are based on kinetic data p and n are stoichiometric coefficients dependent on the cat- with an industrial catalyst at high ( 5 3 mol Vo) inlet butane alyst and are determined by kinetic measurements. concentrations in the temperature range 573-653 K. Only data from isothermal, integral pilot plants 1:1 with commercial Reactor equations reactors and on equilibrated catalyst were used in the final M~~~balance: analyses, the latter point being rarely conceded as important in the literature (Centi et al., 1988; Schneider et al., 1987). c 3 G m dPl ~- - Data from laboratory reactors proved quite misleading in es- - a,?l ( j , R ci) (4) tablishing the final triangular kinetic scheme with product in- (1-WfdZ hibition. This surprising result was independently confirmed (Buchanan, 1985). The temperature and concentration meas- Energy balance: urements along the length of the pilot-plant reactor were crucial for satisfactory parameter estimation, in addition to the pres- ence of maleic anhydride at the inlet to the test section of the dT G CP - = dz c l = J ( - ~ J ) ? l C I ' ) R t i ) ( l - ~ ) - r : r..i.(r--Ts.) o,. reactor to avoid systematic errors. For rigorous testing of the kinetic and reactor models, pre- (5) dictions were compared with data from full-scale pilot-plant where 70) and R U ) are the effectiveness factor and reacrion reactors operated near stability limits. The effects of salt tem- rate, respectively, for the jth reaction and a, is the stoichio- perature, inlet butane concentration and feed rate were illus- metric number for the ith species in jth reactr m , which is trated up to runaway conditions. The heterogeneity of the positive for product and negative for the reactant The pressure model indicated effectiveness factors for the catalyst of about drop across the reactor was calculated using the Er :un equation 0.6 at the hot spot, but little difference between catalyst and (Ergun, 1952). average gas temperatures. Pellet equations: The Reactor Model An appropriate compromise between excessive complexity r" dr r" (Deffl 2) c = 3 J= 1 -aIJR(j) (6) and oversimplification for such highly exothermic fast reac- tions is a one-dimensional, heterogeneous model (Froment, 1972). The utility of such a model has been shown for the :s (Keifr"3 = /=I (7) -(--la,)i?(j) vapor-phase oxidation of benzene to maleic anhydride (Sharma et al., 1984a). The effects of axial and radial dispersion of heat Boundary conditions: and mass can be neglected as a first approximation, since they are relevant only over a short length of the industrial reactor r = O -dT - =dCi = (] (8) in the region of the hot spot. ' dr dr Reaction network A simplified reaction network consistent with the literature (9) and preliminary work is a triangular kinetic model with the desired reaction of butane to maleic anhydride in parallel with an undesired, total oxidation reaction. The decomposition re- action of maleic anhydride to oxidation products closes the network. The stoichiometric equations are: kb hb Bi, = and Bih = - Deif Kerf Desired Reaction The feed entrance conditions to the reactor are bnown. Such Butane + Oxygen Maleic Anhydride + Water equations need to be written only for butane, oxygen, and / maleic anhydride. The corresponding concentrations for CO, Total Oxidation Decomposition Reaction COz, and HzO can be calculated from reaction s: oichiometry Reaction and rates. Using a modified Euler predictor-correc.tor method, the interstitial fluid-phase equations were integrated stepwise CO, CO2, HzO starting at the reactor inlet in conjunction with the pellet equa- Total Oxidation Products tions employing two corrections for optimal usage and a slep 40 January 1991 Vol. 37, No. 1 AI<'hE Journal
- 3. Table 1. Heat transfer Parameters for Butane Oxidation to where Nu is the Nusselt number. The corresponding mass Maleic Anhydride transfer coefficient can be calculated from the heat and mass transfer analogy assuming Prandtl and Schmidt numbers for krsefr = 1.1 W.m-'.K-' air to be equal. Typical ratios for inter- and intraparticle trans- hw*eff = 164-167 W.rn-'.K-' (Bed + inside wall) coefficient = 111-113 W.m-'.K-' port rates are 0.66 for heat and 250 for mass. Overall heat transfer coefficient (Uou) = 105-107 W.rn-*.K-' Biot number for heat = 0.66-0.67 Kinetic model for butane oxidation to maleic anhydride Two types of mechanistic models were considered to describe size of 2 cm (Seinfeld and Lapidus, 1974). The coupled set of the triangular kinetics for the selective oxidation of butane, two-point boundary value equations describing the pellet were i.e., Langmuir-Hinshelwood and Redox mechanisms. Since solved by orthogonal collocation using Jacobi polynomials as alternative kinetic data on a fluidized-bed catalyst of the same expansion functions (Villadsen and Michelsen, 1978). Com- composition showed that, above 15% oxygen in the feed, the putation time for a 3-m-long reactor was about 65 s on a CDC reaction rates were independent of oxygen concentration for 6400 computer. The effectiveness factors 101 for any reaction butane conversions up to lo%, the rate equations could be are computed from the reaction rates integrated over the cat- simplified. Preliminary work in an isothermal integral reactor alyst pellet: had also shown that a maleic anhydride adsorption term was required to better describe concentration profiles in the reactor. The rate equations for the best model were: where a is a geometry factor and equal to 0, 1, and 2 for a R1 = klpy' desired reaction (1 +K2P2) slab, infinite cylinder, and sphere, respectively. decomposition reaction Heat transfer in the reactor tube This has been described by a two-phase continuum model based on a diffusional model for heat transfer that allows for R3 = k3py3 total oxidation reaction (19) different temperature profiles in solid and fluid phases and includes axial conduction in both the fluid and solid (Wellauer To minimize correlation, the rate constants were repara- et al., 1982). The overall heat transfer coefficient is based on meterized as the measured temperature difference between the central axis of the bed and the coolant. It is derived by asymptotic matching of thermal fluxes between one-dimensional and two-dimen- sional models using the effective radial thermal conductivity, kr,eff, the wall heat transfer coefficient, hw,eff. and According where kj673 the value of the rate constant kj at the reference is to the model, the overall heat transfer coefficient U,, can be temperature of 673 K. calculated by: Interpretation of the isothermal data Using the integral method of data analysis, the triangular where al is the smallest root of kinetics leads to the following equations for the system under isothermal conditions and assuming constant density and con- stant number of moles dpl - ( l - ' ) Q ( R , + R , ) Jo is a zero-order Bessel function of the real kind. Table 1 dZ Ft shows typical values of these parameters for one set of ex- perimental data from a full-scale pilot plant. dp2 - ('-'I ( R l - R 2 ) Q dZ Ft Local heat and mass transfer parameters These parameters are essential to calculate surface conditions from measurable gas-phase temperatures and concentrations. The fluid-solid heat transfer coefficient is calculated from the relation (Dwivedi and Upadhyay, 1977) @ - -( 1 - E)Q - [(6-2p)R, + (13-2n)RJ dZ Ft where p I , . . p 4 = dimensionless partial pressures of n-butane, mal- 0.57 eic anhydride, carbon dioxide and carbon mon- = -Pr1'3Rei6Re, 2 50 (16) E oxide, respectively AIChE Journal January 1991 Vol. 37, No. 1 41
- 4. Z = axial bed length Table 2. Tortuosity Factor for Butane Oxidation Catalyst Q = cross-sectional reactor area € = bed voidage Reynolds Peclet Effective ortuosity No. (Rep) No. (Pep) Diffusivit) Factor m*.s-'(x 10") 7 For given rate models, the above equations were integrated - ____ numerically and fitted to observed reactant-product distribu- Nitrogen Pulse tions employing the nonlinear regression program of Klaus and Rippin (1979) utilizing a Marquardt routine. Basic as- 62.2 1.88 1.99 3.3 50.1 1.93 2.15 3 .O sumptions in the program are that there are no systematic 40.8 1.99 2.18 3 .O deviations between the model and the physical system, the 28.4 2.12 2.53 2.6 residuals are only due to errors in the measured responses p 1 19.5 2.31 2.06 3.2 to p 4 , and the errors between different experiments are not 10.7 2.77 2.09 31 correlated. Allowance is made, however, for correlation be- Helium Pulse tween measured responses in any single run. Initial estimates of the parameters were obtained from graphical testing of the 399.9 1.92 2.90 4.9 363.2 1.94 2.61 5.1 data. The objective function S was minimized with respect to 274.6 2.03 2.72 4.8 the parameters g where . P w tracer gases. Since pulse broadening can also be cai sed by axial dispersion in addition to pore diffusion, the formu was meas- where mf is the number of responses in pattern f, nf is the ured independently using identical nonpoi oiis glais particles. number of measurements in pattern f,Pis the number of dif- Table 2 shows the results of the experiments. An ,werage tor- (g) x ferent patterns and Mf is the determinant of the (Mf mf) tuosity of 3 was used for all the species of the react,on mixture, moment matrix of residuals of pattern f. The convergence since the molecular weight of nitrogen is closer to wtane than criterion for each parameter was 1 x helium. Diffusion inside the catalyst Reactor systems used The effective diffusivity of the catalyst was calculated by Kinetic parameters were estimated from experiments carried combining Knudsen and bulk diffusion contributions in the out in isothermal differential and integral reactors both with Bosanquet relation (Satterfield, 1970). crushed and commercial extrudate sizes. The overall experi- mental program is described in Table 3. The diluted bed integral reactor data, from which the kinetic parameters wer 2 ultimately I _-__ - 1 1 - + (26) determined, were fractional in design includiing fc )ur temper- Deff DK,eff DB,eff ature levels, four concentration levels, and ihree flow rates. In a series of 40 experiments, a total of 160 poin:s were ob- where served, and at each point the concentrations of four species were measured. The total of 640 data points were considered sufficient to estimate the minimum expected si., -parameter (27) model. The tortuosity factor r was measured experimentally, 7 is Table 3. Scope of Experiments: Butane Kinetics and Non- the mean pore radius in cm. isothermal Operation ~- Isothermal N misothermal ~- Experimental Studies Laboratory Pilot-Plant 4ot-Plant The catalyst consisted of promoted vanadium-phosphor ox- Differential Low/High High ides whose commercial form is 3 x 10-3-m-dia. extrudates. Converioii Conversion The pore-size distribution by mercury porosimetry showed Total pres., atm 1.1 1-3 1-3 pores in the range 10-s-10-6 m diameter, a pore volume of Salt temp., "C 300-370 320-380 360-400 0.38 x lo-' m3-kg-', and a BET surface area of 1 1 x Total feed rate m2.kg- I . kmolx lO-'h-' 1-3 50-240 40-80 Mole fraction (inlet) C4HIO 0.005-0.03 0.005-0.03 (,018-0.027 Measurement of the tortuosity factor 0 2 0.21 0.21 0.21 A single-pellet string column was used to measure the tor- C4H203 0 0.001-0.002 0 tuosity under nonreacting conditions by pulse broadening Cat. particle size, mm 0.7, I 3 3 Cat. weight, g 1-3 600-740 1,800-2,280 (Cresswell and Orr, 1982). Nitrogen and helium were used as 42 January 1991 Vol. 37, No. 1 AICtiE Journal
- 5. The differential reactor was a 60-cm-long, 35-mm-dia. glass I 1 tube placed in an electrically heated furnace. The temperature in the reactor bed was measured by a sliding chrome-alumel P z I 0 0.7~10-~ m - thermocouple placed in an axial thermowell. Dilution of the 0 7~1O-~rn catalyst with glass beads of the same size in the ratio 1:9 0.56% butane in feed - practically gave isothermal conditions, the temperature vari- ation along the length of the reactor not exceeding 2°C. Butane - conversions did not exceed 8%. Under conditions chosen for the differential reactor, calculations for interphase transport limitations (Gunn, 1978) showed that mass transfer rates were - at least 100 times larger than reaction rates, and gas-solid temperature differences were a maximum of 2°C. cn Y To check the role of product inhibition on kinetics, liquid (394°C) (350°C) (315°C) maleic anhydride was pumped to the inlets of both differential and integral reactors. Practically this proved so difficult that 1.5 1.6 1.7 experiments were carried out only with diluted-bed integral K-' I/T x 1 ~ 3 , reactors to emphasize kinetic measurements in the presence of Figure 1. Influence of pore diffusion on reaction rate. maleic anhydride. In initial work, 3-mm commercial-size ex- trudates were packed in a 4-m-long, 25-mm-ID reactor tube immersed in a stirred molten saltbath for cooling. The tube had several intermediate sampling points and an axial ther- Results and Discussion mowell for temperature measurements. In the front 40% of To study intraparticle diffusion, experiments were carried the reactor, an inert to catalyst ratio of 1:2 was used, concen- out in the differential reactor with 7 x 10-3-m and 0.7 x tration measurements were made only in the second section of 10-3-m-dia. catalyst pellets. The observed rates are shown in the reactor. In a more extensive work using up to 3 mol Vo Figure 1 indicating that pore diffusion is significant above 653 inlet butane, catalyst dilution was used in both sections: 1:l K. Isothermal, integral reactor experiments were, therefore, in the front and 1:0.5 (inert) downstream. performed with 3 x 10-3-m-dia. extrudates operating below The product analyses were performed in the same way as 653 K. the differential reactor experiments. Maleic anhydride in the product gases was absorbed in water and titrated against a Kinetic parameters from the differential reactor standard alkali solution. Butane concentrations at the inlet and outlet were measured using gas chromatography, and CO and Since conversions were small and the inlet concentration of C02 were analyzed by an infrared instrument. Any runs with maleic anhydride was zero, the rate data were first modeled a carbon balance not within *3Vo were rejected. assuming two parallel reactions following power law rate Safety aspects of reactor operation The high inlet butane concentrations in air used in this work were often between the upper and lower flammability limits Table 4. Four-Parameter Simplified Model for Butane of 10.3 and 1.7% by volume, respectively (Perry et al., 1963). Kinetics: Differential Reactor The spontaneous ignition temperature at the stoichiometric kl = (0.94 f 0.73)10-6kmol~kg-1.s~'.atm0.66 concentration of 3.1% is 704 K, which increases to about 820 El = E3 = 63,600 f 13,800 kJ-kmol-' K at 1.7%. Thus salt-bath temperatures of 673 K were not a1 = a3 = 0.66 f 0.16 exceeded in this work for safety reasons. k3 = (0.40 f 0.32)10-6kmol~kg-1.s-1.atm0-66 To further ensure safe operation if flammability limits and Correlation Matrix ignition temperatures were simultaneously reached, standard procedures for such laboratories were additionally used. Rup- k , 1.00 El 0.48 1.00 ture discs at the top of the reactor are large and quick enough k, 0.97 0.47 1.00 to act to mitigate emergencies, automatic hydrocarbon feed 0.99 0.53 -0.97 1.00 shut off with nitrogen purge for variations of pressure and temperature outside predefined limits, visual and audible alarm Objective Function Value = - 288 facilities. The possibility of flame propagation in the fixed catalyst bed is minimal due to the packing acting as a flame arrestor. '>* r l = 30.3% '> r z = 9.6% In the equipment before the packing where the greatest danger lies, entering hydrocarbodair mixtures are preheated only to '> r3 = 24.2% '> r4 = 15.9% about 450 K, well below the spontaneous ignition temperature. Thus, the first few centimeters of catalyst are used to bring Stoichiometric Coefficient: n = 5.5 the feed up to reaction temperatures, but safely inside the t - - (standard deviation of residuals of fitted response/range of packed bed. In equipment after the packing, temperatures drop response) x 100 1 = butane; 2 = maleic anhydride; 3 = C 0 2 ;4 = CO sharply, while exit concentrations of butane are well below f = 95% confidence limits flammability limits in high conversion operation. AIChE Journal January 1991 Vol. 37, No. 1 43
- 6. expressions. This six-parameter model showed that k3 and a3 Table 5. Seven-Parameter Model for Butane Kinetics: were poorly determined and with a strong correlation between Isothermal Integral Reactor rate constant and reaction order. A simplified four-parameter model gave better estimates, which are shown in Table 4 with kl = (0.96 f 0.28)10-6kmol.kg-1.s-' .atm0.54 El = E3 = 93,100 f 5,700 kJ.mol-I their single-parameter 95% confidence limits, correlation ma- a1 = a3 = 0.54 f 0.05 trix, objective function value and accuracy of fit. Joint pa- k2 = (0.29 f 0.14)10-5kmol-kg~1~s-1~atm-' rameter confidence limits are expected to be somewhat larger, E2 = 155 f 35 MJ.kmol-I about 50%. However, reducing the total number of parameters k, = (0.15 f 0.03)10-6 k m o l . g - ' . ~ - ' ~ a t m - ~ . ~ ~ K2 = 310 f 125 atm-' from six to four. did not increase the objective function sig- nificantly. Also, kl and k3 are highly correlated as are kl and Correlation Matrix al, k3and a ] .The inability to control the maleic anhydride and k, 1.00 feed to the inlet of the differential reactor probably led to El 0.68 1.00 systematic errors accompanying the differential data (Cropley, k3 0.87 0.55 1.00 1987). The estimate of 63,600 f 13,800 kJ/kmol-' is about a1 0.94 0.62 0.94 1.00 50% lower than subsequent integral data which included prod- kz 0.50 0.46 0.08 0.34 1.00 E2 -0.20 -0.31 0.09 -0.13 -0.83 1.00 uct inhibition at the reactor inlet. K2 0.69 0.63 0.39 0.44 0.61 -0.22 1 . M Incorporating the kinetics from Table 4 into the reactor model (Eqs. 4-1 1) and comparison with experimentally ob- Objective Function Value = - 606 served temperature and concentration profiles from a 4-m pilot- plant reactor indicated severe discrepancies. Since reaction rates were underpredicted in the front end and overestimated in the tail end of the reactor, the kinetic model was modified to include inhibition by the product maleic anhydride. This is '> r3 = 6.8% '> r 4 = 2.3% consistent with an in situ FTIR study of n-butane selective oxidation to maleic anhydride on V-P-0 catalysts in which the Stoichiometric Coefficients: p =I, n = 5.5 carbonyl stretching vibrations of maleic acid and maleic an- 8 - (standard deviation of residuals of fitted responsehange of - hydride were observed (Wenig and Schrader, 1986). response) x 100 1 = butane; 2 = maleic anhydride; 3 = CO,; 4 = CO f = 95% confidence limits Kinetic parameters from isothermal integral reactors The initial integral reactor work in the 4-m tube, with only the front end diluted, was combined with independent meas- urements of maleic anhydride decomposition kinetics in a lab- ing the decomposition reaction. The resulting seven-parameter oratory salt-bath reactor (Kuhn, 1979). The latter work led to model, Table 5, gave better estimates, decreased the objective the use of incorrect stoichiometryp=2, n = 5.5, when applied function value, and decreased the 070 standard deviation in to full-scale pilot-plant data where CO/C02 ratios were always comparison to Table 4. A maleic anhydride adsorption term greater then 1.0. The more extensive work using up to 3 mol is now included, and the activation energy of the decomposition '70 inlet butane with catalyst dilution in both sections was, reaction, 155 MJ.kmol-' compared to the desired reaction therefore, used to determine all the kinetic parameters includ- 93.1 M J -kmol-I, clearly shows the disadvantage with respect Table 6. Testing of the Butane Reactor Model against Pilot-Plant Data Days On Stream 111 112 113 114 115 Operating Conditions 1.67 1.67 1.27 0.95 0.95 Feed rate, m3.h-' Inlet butane conc., Vo 1.81 1.81 2.20 2.57 2.68 Salt temperature, "C 383 363 363 363 373 Experimental Data 2 600 Hot spot, "C 420 375 380-85 390-95 Runaway! Conversion, '70 85 60 65 50 60-65 Selectivity*, 9 ' 0 55 67 63 56 - Model Predictions Runaway Hot spot', "C 422 383 390 40 1 (Salt temp. = 376°C) Conversion, T o 82.3 58.6 64.4 72.9 92 Selectivity, Vo 61.1 66.4 63.5 59.3 53 CO/C02 ratio 1.48 1.25 1.26 1.27 1.62 Overall heat trans. coeff. 106 104 93 83 85 UOu, W.m-2.K-1 'Hot spot refers to axial gas temperature **selectivity = moles maleic anhydride, ma/mol.ma + mol.(CO+ C02)/4. 44 January 1991 Vol. 37, No. 1 AIChE Journal
- 7. A Salt temp. 400°C Inlet C4H,,% 1.82 Feed rate, m3/h 1.69 t ,, t I "o I Reactor diameter = 0.024m Feed flow rate = 0 96 kg m+ s - I l6 0 390°C 1.86 1.68 VHSV = 703 h-' 380°C 0.75 1.65 0 450 fi 0 v 370°C 350°C 0.75 1.26 1.65 1.66 .m E 1 x"3 0 extrudate 6 -2 2 430 1.0 0 410 0.8 0" E N I 390 d 0.6 0" 8 1 370 0.4 - 2 v v v v 350 0.4 0.8 12 1.6 2.0 24 2.8 3.2 3.6 4.0 4.4 4.8 0.2 Reactor length, m Figure 2. Maleic anhydride exbutane temperature pro- 0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 Reactor length, rn files at various operating conditions: experi- mental catalyst Figure 3. Observed and predicted reactant, product and temperature distributions in a full-scale re- actor tube. to selectivity of operation with high hot spots. A temperature dependence of K2 could not be estimated with any certainty so it was considered constant. days on stream, predicted conversion is much higher than The relative contribution of reactions 2 and 3 to the for- observed due possibly to an observed reactor upset since op- mation of CO and COz is reflected in the new values of the eration was on the verge of stability. Runaway was predicted stoichiometric coefficients p = 1, n = 5.5. These values allow at a salt temperature 3°C higher than observed at a butane CO/C02 ratios greater than 1.0 as frequently seen in non- inlet concentration of 2.68%. isothermal operation with hot spots. In a series of runs, the model was further tested using an experimental catalyst. The temperature profiles at the respec- Model predictions vs. pilot-plant data tive conditions are shown in Figure 2 and the comparison of For rigorous testing of the kinetic and reactor model, pre- experimental and predicted exit concentrations are shown in dictions were compared with data from a full-scale pilot-plant Table 7 . For salt temperatures between 350-4OO0C, Figure 2 reactor operated near stability limits due to low feed rateJ and shows that temperatures are well predicted. Table 7 shows that high hydrocarbon concentrations. The reactor was 25-mm-dia, conversion and selectivity are satisfactorily predicted especially 5-m-long, and fitted with an axial thermowell. The catalyst at higher conversions. was very similar to that used for kinetic measurements. Table Observed and predicted concentration and temperature pro- 6 shows a comparison of model predictions vs. experimental files in a full-scale reactor tube are shown in Figure 3. The data at different operating conditions. Conversion and selec- conversion of butane and selectivity to maleic anhydride are tivity have been calculated at average gas temperatures given well predicted. The hot-spot temperature is underpredicted but by the one-dimensional model while predicted hot-spot tem- within the uncertainty of the heat transfer coefficient (f10%) peratures are for the gas at the axial location, using the two- (Wellauer et al., 1982). The combined CO + C 0 2 profile dimensional continuum model of heat transfer, as described illustrates the relative contributions of reactions 2 and 3 and earlier. The data from 111, 112 and 113 days on stream indicate tests the stoichiometry shown in Table 5 , p = 1, n = 5 . 5 . The that the model satisfactorily predicts the effects of salt tem- CO/C02 ratio at the reactor exit as shown in Table 7 also tests perature, inlet butane concentration, and feed rate. At 114 the suitability of the chosen stoichiometry. Table 7. Testing the Model with Data from an Experimental Catalyst in Pilot-Plant Reactors Conditions 1 2 3 4 5 Operating Conditions Feed rate, m3.h-' 1.68 1.68 16 .5 1.65 1.66 Inlet butane conc., Vo 1.82 1.86 0.75 0.75 1.26 Salt temperature, "C 400 390 370 380 350 Inlet pressure, atm 19 . 1.65 1.64 1.64 1.62 Experimental/Predicted Hot spot, "C 440/451 419/424 3801380 392/393 360/359 Conversion, % 90/90.8 84/77.2 81 /72.4 89.4/82.4 52.3/42.2 Selectivity, Yo 60/57.5 68/63.4 71.9/73.2 66.1/71.2 81.6/74.2 CO/C02 ratio 1.51/1.7 1.55/1.49 1.56h.37 1.57/1.51 1.43A.16 Pressure drop, atm 0.9/0.6 0.65/0.61 0.64/0.70 0.64/0.72 0.62/0.59 Heat transfer coeff., W.rn-'.K-' /lo7 /lo4 /lo6 /lo7 /lo2 AIChE Journal January 1991 Vol. 37, No. 1 45
- 8. Table 8. Heterogeneity of the Reactor Model Operating Conditions Exp . Pied. _____ Feed rate = 1.68 rn3.h-' Hot-spot temp. 440°C 45l'C Salt-bath temp. = 400°C Conversion 90% 90.8% Inlet butane = 1.82% Selectivity 60Vo 5E VO - ____ Reactor Temperature, "C Effectiveness Factor Selectivity (- O/CO,> Location Avg. Axial Pellet 91 92 73 m Surface 0.2 411.5 415.0 413.3 0.62 2.80 0.98 77.1 1.20 0.4 436.5 447.7 439.1 0.66 1.72 0.95 71.9 1.34 0.6 438.2 447.9 440.6 0.72 1.38 0.94 68.8 1.45 1.o 423.9 431.2 425.3 0.82 1.15 0.95 65.8 1.51 2.0 412.4 416.2 413.1 0.89 1.06 0.95 62.4 1.56 4.0 406.1 408.0 406.5 0.90 1.02 0.92 58.8 1.65 - __-_ The heterogeneity of the model is illustrated from data with Notation a fresh catalyst at a salt temperature of 400°C. Table 8 shows u,, = number of moles of species I invohed in reaction J that in the hot-spot region, the pellet surface is about 2-3°C A = reactor surface areaheactor volumc., m higher than the average gas temperature. Intraparticle tem- b = half thickness of the active phase iri the ~tellet,m perature gradients are negligible. By assuming a parabolic ra- C,= concentration of species I , mo1.m ' C, = specific heat of the gas, J .kg-'.K dial temperature profile, it is possible to estimate that, in the dp = diameter of sphere of equal volume to surface area, m hot-spot region, the average gas temperature in the tube cross- D = axial dispersion coefficient, m2.s-' , section is about 11"C lower than the axial value predicted by D,,, = effective diffusivity of species I , m z 3 - l the model. E, = activation energy of reaction step], .I.mc I - ' E, = catalyst particle porosity Figure 4 illustrates the region of stable operation in contrast Ft = total gas molar flow rate, mo1.s to runaway, with feed rate as a parameter. The development G, = gas mass velocity, kg.m-*.s-' of more active and selective catalysts would allow inlet butane hw,erf= apparent wall heat transfer coefficient, I .m2.K-' concentrations of 4 5 % to be used in stable operation in axial- -m, = heat of reaction of step j , J.mol ' kg, h = interphase mass and heat transfer ( oefficie, it, m . s- I and flow fixed-bed reactors. W .m-2. K-' kgs = molecular conductivity of air, W m-'.K I k, = rate constant of step j kreff = effective radial thermal conductivity )f the bed, Acknowledgment W.m-l .K-' One of the authors (RKS) acknowledges financial support from K,,, = effective thermal diffusivity of the catalyst pellet, KWF Bern, ETH Zurich and Swiss Aluminium, Neuhausen. The re- W .m-' .K-' actor work was performed by Dr. J. P. Stringaro, Messrs. M. Bollinger K2 = maleic anhydride adsorption rate constanr atm-' and H. P. Keller in Neuhausen. This article was presented in prelim- M = molecular weight inary form at the AIChE Meeting in San Francisco, November 1984, n, p = number of moles of oxygen required/molt of reactant the delay in publication being a requirement of a confidentiality agree- Pep = Peclet number of axial dispersion (ullrp/D6Tx) ment. p , = mole fraction of species I in the mixture Pr, Sc = Prandtl and Schmidt number for the gas R = universal gas constant, J.mo1-I.k r = distance from pellet center to any point insitie the pellet, m 390 I I I I I I R U ) = intrinsic reaction rate of j , rno1.s- 'in-3 Re, = Reynolds number (G,dp/p) 0 380 Runaway region - S = objective function defined in Eq. 25 T , P = temperature and pressure, K and atm U,,, = overall heat transfer coefficient, W . n ~ - ~ . h -' u = linear gas velocity, m.s-l E 0. Z = distance from reactor inlet, m 5 L 360 -1 Stable operation Greek letters $ 350 - (feed rate 0.95nm3/h) 2-Dim. model p = gas density, k g . r ~ - ~ 70') = effectiveness factor for step j 340 I I I I I I I E = catalyst bed voidage 1.4 1.8 2.2 2.6 3.0 3.4 3.8 4.2 4.6 p = gas viscosity, kg.m-l.s-' Inlet mole fraction x 1 0 2 , % 7 = catalyst tortuosity factor 01 U) = exponent in rate equation for step .i Figure 4. Region of 'STABLE' operation. Cl = cross-sectional area of reactor tube. niL 46 January 1991 Vol. 37, No. 1 AIChE Journal
- 9. Subscripts Single and Multiresponse Situations,” Comp. and Chem. Eng., 3, 105 (1979). 1, 2, 3, 4 = butane, maleic anhydride, CO,, CO Kuhn, P., “Maleic Anhydride Decomposition,” internal report, Alu- K, B = Knudsen, bulk suisse (1979). G , MA = gas, maleic anhydride Malow, M., “Benzene or Butane for MAN,” Hydro. Processing, 149 0, p = bulk gas or reactor inlet, particle (Nov., 1980). s, t = salt, tube Neri, A., and S. Sanchioni, US 4.314.946 to Ftalital S. p. A., “Process for the Continuous Separation of Maleic Anhydride from Process Gases,” (Feb. 9, 1982). Literature Cited Perry, R. H., C. H. Chilton, and S. D. Kirkpatrick, Chemical En- Buchanan, J. S., private communication (Mar. 1985). gineers’ Handbook, 4th ed., McGraw Hill, New York, 9 (1963). Budi, F., A. Neri, and G . Stefani, “Future MA Keys to Butane,” Satterfield, C. N., “Mass Transfer in Heterogeneous Catalysis,” Chap. Hydro. Proc., 159 (Jan. 1982). 1, MIT Press, Cambridge, MA (1970). Centi, G . , F. Trifiro, J. R. Ebner, and V. M. Franchetti, “Mechanistic Schaffel, G . S., S. S. Chen and J. J. Graham, “Maleic Anhydride Aspects of Maleic Anhydride Synthesis from C4 Hydrocarbons over from Butane-Catalytic Oxidation in a Fluidized Bed,” Erdoel und Phosphorus Vanadium Oxide,” Chem. Rev., 88, 55 (1988). Kohle, Bd. 36, Heft 2, 85 (1983). Contractor, R. M., H. E. Bergna, H. S. Horowitz, C. M. Blackstone, Schneider, P., G. Emig, and H. Hofmann, “Kinetic Investigation and U. Chowdhry, and A. W. Sleight, “Butane Oxidation to Maleic Reactor Simulation for the Catalytic Gas-Phase Oxidation of n- Anhydride in a Recirculating Solids Reactor,” Catalysis 1987, 647, Butane to Maleic Anhydride,” Znd. Eng. Chern. Res., 26, 2236 J. W. Ward, ed., Elsevier, Amsterdam (1988). (1987). Cresswell, D. L., and N. H. Orr, “Measurement of Binary Gaseous Seinfeld, J. H., and L. Lapidus, “Process Modelling, Estimation and Diffusion Coefficients within Porous Catalysts,” Residence Time Identification,” Vol. 3, Mathematical Methods in Chemical Engi- Distribution Theory, G . PethaandR. D. Nobel, eds., Verlag Chemie neering, Prentice Hall (1974). GmbH, D-6940 Weinheim, 41 (1982). Sharma, R. K., D. L. Cresswell, and E. J. Newson, “Selective Oxi- Cropley, J. B., “Systematic Errors in Recycle Reactor Kinetic Stud- dation of Benzene to Maleic Anhydride at Commercially Relevant ies,’’ Chem. Eng. Prog., 83, 46 (1987). Conditions,” Znt. Chem. Eng. Symp. Ser., No. 87, Edinburgh (Sept., De Maio, D. A., “Will Butane Replace Benzene As a Feedstock for 1984a). Maleic Anhydride?” Chem. Eng., 104, (May 19, 1980). Sharma, R. K., D. L. Cresswell, and E. J. Newson, “Kinetics and Dwivedi, P. N., and S. N. Upadhyay, “Particle-Fluid Mass Transfer Reactor Modeling in Fixed-Bed Pilot-Plant Production of Maleic in Fixed and Fluidized Beds,” Znd. Eng. Chem. Proc. Des. Dev., Anhydride by the Oxidation of n-Butane,” AIChE Meeting, San 16, 157 (1977). Francisco, paper 69b (Nov. 1984b). Ergun, S., “Fluid Flow through Packed Columns,” Chem. Eng. Prog., Villadsen, J., and M. L. Michelsen, Solution of Differential Equation 48, 89 (1952). Models by Polynomial Approximation, Prentice Hall (1978). Escardino, A., C. Sola, and F. Ruiz, Znd. Eng. Chem. Prod. Res. Wellauer, T., D. L. Cresswell, and E. J. Newson, “Heat Transfer in Dev., 18, 7 (1979). Packed Reactor Tubes Suitable for Selective Oxidation,” ACS Symp. Froment, G . F., “Fixed-Bed Reactors-Steady-State Conditions,” Ser., 196, 527 (1982). Proc. Znt. Sym. Chem. Reaction Eng., Elsevier, Amsterdam (1972). Wellauer, T., D. L. Cresswell, and E. J. Newson, “Optimal Policies G u m , D. J., “Transfer of Heat or Mass to Particles in Fixed and in Maleic Anhydride Production through Detailed Reactor Mod- Fluidized Beds,” Znt. J. Heat and Mass Transfer, 21, 467 (1978). eling,” Chem. Eng. Sci., 41, 765 (1986). Hodnett, B. K., “An Overview of Recent Developments in Elucidating Wenig, R. W., and G . L. Schrader, “Zn Situ FTIR Study of n-Butane the Mechanism of Selective Oxidation of C-4 Hydrocarbons over Selective Oxidation to Maleic Anhydride on V-P-0 Catalysts,” J. Vanadium Phosphorus Oxide Catalysts,” Cat. Today, 1,527 (1987). Phys. Chem., 90, 6480 (1986). Klaus, R., and D. W. T. Rippin, “A New Flexible and Easy to Use General Purpose Regression Program for Handling a Variety of Manuscript received Sept. 11, 1989, and revision received Nov. 5, 1990. AIChE Journal Januarv 1991 Vol. 37. No. 1 47