maleic anhyride production method

2652 Ind. Eng. Chem. Res. 1992,31, 2652-2660

Modeling the Catalytic Oxidation of n-Butane to Maleic Anhydride in a Circulating Fluidized Bed Reactor

Todd S. Pugsley,t Gregory S. Patience,*?* Franco Bermti,* and Jamal Chaoukif Department of Chemical and Petroleum Engineering, University of- Calgary, 2500 University Dr. NW, Calgary, Alberta, Canada T2N lN4, and Departement de Genie Chimique, Ecole Polytechnique de Montreal, C.P. 6079, Succursale A, Montreal, Quebec, Canada H3C 3A7

A circulating fluidized bed (CFB) is a gas-solid contactor in which solids are transported vertically in a riser by a high-velocity gas stream. CFB reactors are becoming increasingly attractive for conducting specific catalytic reactions. The catalytic oxidation of n-butane to maleic anhydride (MAN) has been simulated for a ao-m-high, 30-cm-diameter CFB to investigate the reactor performance under various operating conditions. The comprehensive simulation combines a core-annular hydrodynamic model of the CFB with a fixed bed reaction kinetics model. Reactor performance is evaluated in terms of n-butane conversions, product yields, and selectivity. The simulation can easily accommodate different reaction kinetics and therefore can be used to predict the performance of a CFB reactor for any catalytic process.

1. Introduction Circulating fluidized bed (CFB) reactors are typified by

riser superficial gas velocities normally in the range of 2-10 m/s, compared with less than 1 m/s in conventional bubbling beds. At these gas velocities, solids throughput is very large while reactor diameter is minimized and gas residence time is short. Solids carried out of the riser are separated from the gas by cyclones and are returned to the bottom of the riser by a vertical standpipe.

The traditional uses of a CFB have been mainly in noncatalytic proteases, such as the combustion of low-grade fossil fuels. However, the unique design of a CFB also makes it attractive as a catalytic reactor. Figure 1 is a schematic of a CFB catalytic reactor for the partial oxidation of n-butane to maleic anhydride. Solid catalyst particles are injected from the standpipe and are transported vertically in the high-velocity gas stream containing n-butane. The reaction occurs along the riser length, and the catalyst is rapidly deactivated. The solids and gas exit from the riser via a smooth elbow and are separated in the cyclone. The product stream containing maleic anhydride and other gases is sent downstream for purification. The deactivated catalyst particles are stripped of any carbo- naceous material by an inert such as nitrogen or steam and sent on to the regenerator section where contact with air oxidizes the catalyst surface once again. The regenerated catalyst is gravity fed to the standpipe where it moves to the base of the riser for reinjection into the riser reaction zone.

Contractor and Sleight (1987) discussed the inherent advantages of a CFB in the catalytic partial oxidation of n-butane to maleic anhydride (MAN). These include separate catalyst reduction and oxidation zones in the riser and standpipe, respectively, low catalyst inventory, and uniform temperature throughout the riser (elimination of hot spots). Since the catalyst may be regenerated in a separate vessel, oxygen requirements in the feed are minimal, thus leading to high selectivity to MAN and a more concentrated product stream. Contractor and Chaouki (1990) reviewed other potential uses for the CFB as a catalytic reactor.

* To whom correspondence should be addressed. University of Calgary. t Ecole Polytechnique de Montreal. *Present address: E. I. du Pont de Nemours & Co., Wilmington,

DE.

The interest in the partial oxidation of n-butane to maleic anhydride is due to the importance of the reaction on an industrial scale. For the first 40 years of maleic anhydride production, the feedstock was benzene (DeMaio, 19801, but due to increasing environmental concerns over the use of benzene and escalating benzene costa, almost all US. plants have switched to n-butane. Annual MAN production in the United States is approaching 19OOOO metric tons and is increasing by approximately 6% per year to meet increasing demand (Irving-Monshaw and Kislin, 1989). MAN is used primarily in the production of unsaturated polyester resins, alkyd resins, and specialty copolymers. I t is also used in the manufacture of agri- cultural chemicals, as an additive in lubricating oils, and as a building block for L-aspartic acid, used in making NutraSweet. The high demand for MAN combined with the interest in the circulating fluidized bed technology for its production makes a comprehensive simulation for predicing reactor performance a valuable tool for industry.

2. Kinetics 2.1. Reaction Kinetics Rate Equations. Intrinsic rate

equations and reaction pathways for the partial oxidation of n-butane to maleic anhydride (MAN) on a vanadium phosphorus oxide (VPO) catalyst utilizing a packed bed experimental reactor have been proposed by Centi et al. (1985). These expressions were later cited in the work by Contractor and Sleight (1987) and Patience and Chaouki (1990). The complete reaction pathway involves both series and parallel reaction steps:

(1)

n-C4Hlo - C 0 2 (2) n-C4Hio - MAN - COZ The rate equations proposed by Centi et al. (1985) are

(3)

r2 = rCOs = kZc,B

r3 = -rMAN = k$MAN(CoY/CB6)

(4)

(5)

where rMAN is the rate of MAN formation from n-butane, is the rate of C02 formation from n-butane, and -rm

?&e rate of maleic anhydride decomposition to C02 and water. The kinetic parameters determined by Centi et al. (1985) are summarized in Table I.

0888-5885/92/2631-2652$03.00/0 0 1992 American Chemical Society

Ind. Eng. Chem. Res., Vol. 91, No. 12,1992 2653

proportional to the number of catalytic sites st i l l active. If q(t) corresponds to the number of deactivated sites at solids residence time t., and qo to the maximum number of sites for a totally deactivated catalyst, then the rate of maleic anhydride (MAN) formation, for example, h o m e s

RISER

/-MALEIC ANHYDRIDE

7 OFT GAS

REGENERATOR

AIR

STANDPIPE

BUTANE

Figure 1. Schematic of a eirculatiag fluidized bed an a catalytic reactur.

Table I. ginetia Param&- for Itate Quntions of Centi et

340 6.230 9.040 0.966 KB = 2616 mol/L; (1 - @ = 0.2298; 7 = 0.6345; 6 = 1.151. Unlike the resulta of other studies (Wohlfaiut and

Hoffman, 1980, Sharma et al, 1991), the catalyst employed by Centi and mworkera was very selective toward maleic anhydride. The vanadium present in the VPO catalyst is the main oxygen carrier, and during the reaction with the hydrocarbon in the gas phase, certain crystalline phases in the solid catalyst are transformed via a reduction pro- ceea. This is written in terms of the valence change as V+5 - V+. Contractor (1988) found the average oxidation state of the catalyst to he about 4.1, with a difference of 0.2 in the average oxidation state between the oxidized and reduced state of the catalyst. This indicates that most of the active catalyst exists in the V4 oxidation state and that the catalyst surface provides most of the oxygen needed for the reaction. As discussed by Centi et al. (1988). the kinetic model of Centi et al. (1985) is indicative of a fresh catalyst poeaessing large amounts of surface oxygen and readily forming the V+ crystalline phase. Therefore, the model of Centi et al. (1985) is appropriate to use for the modeling of the CFB reador because of the continuous reiniection of reeenerated &e.. fresh) catalvst from the

~

vertical standpiie. 22. Catalyst Deactivation Model. The reduction of

the VPO catalvst is an examole of oarallel deactivation. where the rea&& form prdrxta &d, at the same time; deactivate the catalyst. The rate of reaction then becomes

This parallel deactivation is repreaentsd by Smith (1981) as a simple fmt-order reaction at the catalyst surfam, and thus the rate of catalyst deactivation, rd, is writtan as

Integration of eq 7 gives

Subetitution of eq 8 into eq 6 provides the rate of reaction incorporating catalyst deactivation:

b ( t 3 = MAN exp[-at.l (9) The solids residence time used in the catalyst deactivation model is calculated on the basis of the upward solids v e locity in the core region. Aa a result, it is assumed that the catalyst deactivation in the core and annular regions is the same. This is a large simplifying assumption because the solids present in the annular region will have been in the riser for a longer period of time due to solids recirculation down the riser walL Furthermore, the solids present at any point in the riser will have a varied history due to solids crossflow between the core and annular regions. A more rigorous formulation would incorporate the solids residence time distribution (RTD), and more work is required in this area to improve the simulation. The purpose of this simplified deactivation model was to acknowledge the deactivation, which is known to be very fast, and to investigate the importance of the deactivation on reactor performance.

3. Riser Hydrodynamics 3.1. Riser Flow Structure. The flow pattam of the

solids in the riser of a CFB is extremely complex and is dependent on many variables. The column diameter, height, and exit configuration, particle properties, and gas characteristics all affect the flow structure in the riser. Visual observation in a riser of circular cross section indicates the existence of a lean suspension of solids in gas flowing upward in the center of the riser, with a denser downilow of solids next to the wall. Weinsteii et aL (1986) and Bader et al. (1988) both observed decreasing radial voidage profdes in their experimental units, with the most dramatic decrease occurring very close to the wall. The radial location of this sudden change in voidage indicates the interface between the dilute upward flowing region and the dense downward flowing region. The CFB riser is also characterized by a dense, turbulent portion at ita base where the solids are introduced into the riser from the standpipe, which becomes leaner as the flow of solids d e velops and the particles are accelerated to their steady- state upward velocity in the riser. 3.2. Berruti-Kalogerakis Model. Bermti and Ka-

lcgerakis (1989) modeled the CFB riser as behg characterized by a ooreannular type of flow stmctwe, with solids moving upward in the lean core region and downward in the dense annular regior.. The model assumes that the solids move dmnward ii the dense annular region at the

2654 Ind. Eng. Chem. Res., Vol. 31, No. 12,1992

terminal velocity of a single particle, and that the voidage in the dense annular region is the same as that of a bubbling bed of the solids at minimum fluidization. This model requires the solids circulation rate, the axial superficial gas velocity, and the average axial riser voidage profile as inputs.

The average axial voidage profile can be obtained from an experimentally determined pressure drop profile ac- cording to

1 d P eaVg = 1 - - -

P& dx Equation 10 assumes that the axial pressure drop profile is due solely to the weight of solids present at any axial location, ignoring frictional effects at the wall. Arena et al. (1986) compared the average axial voidage profile predicted from pressure drop measurement with the profile obtained from trapped solids weighing along the riser length. Pressure drop measurements were found to underpredict the average voidage in the riser, and this dis- crepancy was attributed to the frictional effects of the wall. They indicated that the small diameter column used (4.1-cm i.d.) may have enhanced the frictional effects.

Berruti and Kalogerakis (1989) derived the following expression for the voidage in the core region of the riser:

A solids balance over a volume element of the riser resulted in the following expression for the core radius, r,:

Equations 11 and 12 may be solved simultaneously to obtain the values of core voidage and core radius at any axial position along the riser. The limitation of the Berruti and Kalogerakis model is the requirement of an experimentally determined axial pressure drop profile. There- fore, the model can only be applied to experimental unita for which the pressure drop profile has been obtained.

3.3. Predictive Model of Wong et al. (1992). Wong et al. (1992) have recently developed a predictive model for the average axial riser voidage profile which is used in conjunction with the Berruti-Kalogerakis model to describe the internal flow structure of a CFB. Given riser geometry, solids and gas physical properties, solids circulation rate, and superficial gas velocity, the average axial voidage profile may be calculated along with the core voidage and core radius, core gas and solids velocity, and solids interchange coefficienta between the core and annular regions, all functions of axial location along the riser.

This model considers the CFB riser divided into three sections: an acceleration zone corresponding to the dense region at the base of the riser, a fully developed flow zone immediately above the acceleration zone, and a deceleration zone for risers equipped with an abrupt exit geometry. If the riser has a smooth exit to the cyclone, the fully developed zone is assumed to extend all the way to the outlet.

3.3.1. Acceleration Zone. In the dense region at the base of the riser of a CFB, a fraction of the total measured pressure drop between the entrance and the start of the fully developed flow zone can be attributed to the acceleration of the particles to their steady-state velocity. If

this acceleration component is not separated from the pressure drop data, erroneous voidages will be predicted in the acceleration zone. The voidage predicted from pressure drop data if the acceleration component is not separated is referred to aa the apparent voidage, ewp The voidage predicted once the acceleration component is separated is called the actual voidage. The acceleration zone is modeled assuming a predominantly upward flowing gas-solids suspension. Wong et al. (1992) derived the following expression for the aparent solids holdup in the acceleration zone:

where 1 - e, r=1- - (14) 1 - 0

The first term on the right-hand side of eq 13 representa the effect of the acceleration of particles on the observed preaeure drop, while the gecond term on the rightrhand side represents the effect of the weight of solids.

In order to evaluate the constant r, the apparent voidage at the bottom of the riser, q,, must be known. Wong and co-workers proposed an empirical correlation for eb developed from a regression analysis on a large pool of published experimental data. For Geldart group A solids the correlation is

-0.0 2 6 2 8 ~ ~ - 0 . 0 7 9 4 R ~ -0.12016 (15) eb = 0.714( '.) Ps uo P

This value is substituted into eq 13 with x = 0. The solids velocity gradient in the acceleration zone is assumed to be proportional to the difference between the steady-state solids velocity in the fully developed flow zone, U,,, and the average solids velocity in the acceleration zone, U,:

(16)

Assuming U, is zero at the riser inlet ( x = 0), and taking U, as 0.99U8, at the end of the acceleration zone (where x = LA, then the proportionality constant k is determined from the integrated form of eq 16 with the substituted limits:

k = -h(O.Ol/Lacc) (17) Thus, if the length of the acceleration zone, Lam, is known, a value of the proportionality constant, k, may be determined. Wong et al. (1992) used a modified version of the correlation of Enick et al. (1987) to calculate the length of the acceleration zone:

(18) 3.36. Developed Flow Zone. Immediately above the

acceleration zone, a fully developed flow zone is assumed where the solids and gas have been accelerated to their steady-etate velocity, and the average solids holdup re- mains essentially constant.

Patience et al. (1993) have found that in the fully developed flow regions of circulating fluidized beds, the slip factor, defined as the ratio of the interstitial gas velocity to the solids velocity, is approximately equal to 2:

Ind. Eng. Chem. Res., Vol. 31, No. 12, 1992 2655

Knowing that the average solids velocity in the fully developed zone is given by

(20)

eq 19 may be rearranged to solve for the average voidage in the fully developed flow zone:

Equations 13 and 21 will fully describe the average axial voidage profdes in CFB risers equipped with a smooth exit configuration. Typically, average voidage values in the range of 0.85-0.98 are observed in CFB risers. As shown in Figure 1, the CFB catalytic reactor for the partial oxidation of n-butane is equipped with a smooth exit, and therefore this simulation of the CFB catalytic reactor does not incorporate the deceleration zone.

3.3.3. Solids Interchange Coefficients. The model of Wong et al. (1992) uses the empirical expression of Bolton and Davidson (1988), based on a turbulent diffusion mechanism, to calculate the solids interchange coefficient between the core and the annular region:

1 - 2.0Re-lI8 1 + s / 1 2 K, = 0.1a1/2Uo

The determination of the annulus-to-core solids interchange coefficient is based on an analogy with gas-liquid countercurrent annular flow. It has been suggested (Senior and Brereton, 1990; Takeuchi and Hirama, 1990) that the diffuse interface between the core and annular regions of a CFB is comparable to that of a gas-liquid annular flow reported in the literature. The expression for the solids annulus-to-core solids interchange coefficient is

where pmm and Ucom are the combined phase core density and the combined phase core velocity, respectively, and defined by Wong et al. (1992). The quantity fi, is the smooth pipe friction factor. The proportionality constant, 4, is determined for the case of a negligible solids wall layer thickness such that r, = R and the net solids interchange is mro. The detailed procedure for calculating 4 is outlined by Senior and Brereton (1990).

The simulation presented here predicts core-to-annulus solids interchange coefficients in the range of 1-3 m/s, while the predicted annulus-to-core solids interchange coefficients vary between 0.2 and 1 m/s. The largest values of the solids interchange coefficients occur in the dense, turbulent bed at the riser base and then decrease as the fully developed flow zone forms.

3.3.4. Gas Interchange Coefficients. In this work, the Higbie penetration theory (Higbie, 1935) is used to calculate the gas interchange coefficient. The theory was originaUy developed to describe contact between liquid and gas occurring for a short period of time. Thus a steady- state concentration gradient would not have time to de- velop as the gas moves into the liquid. This results in the following expression for the gas mass-transfer coefficient:

k, = (4Dmt/~t)'/~ (24) Eaending this to the CFB, the crossflow of gas is assumed to occur in the coreteannulus direction only. The voidage in eq 24 is the annular voidage, and the diffusivity Dm is

Table 11. Summary of the Major Assumptions Used for the Computer Simulation the CFB catalytic reactor operates isothermally chemical reaction occura in both the core and annular regions catalyst deactivation i a the same in both core and annular

gas input to the annular region is due solely to crossflow of gas

the chemical reaction is kinetically controlled rieer diameter = 0.3 m; riser height = 20 m riser is equipped with a smooth exit to the cyclone operating conditions G, = -800 kg/(m2 8) Uo = 4-6 m/s CB = 1-50 mol % T = 573 K

D, = 75 pm p, = 1500 kg/m3

Ut = 0.05 m/s

regions at the same riser axial location

from the core

W O catalyst physical properties (Geldart A)

e d = 0.5

that of n-butane in air. The predicted values of the gas interchange coefficients are of the order of m/s, while data supplied by G. S. Patience calculated from an experimental gas RTD indicate values for the core-bannulus gas interchange coefficient in the range of 0.015-0.094 m/s. Thus there is agreement between the simulation and the experimental data.

4. Computer Simulation The major assumptions incorporated into the computer

simulation are outlined in Table 11. The CFB catalytic reactor is assumed to operate isothermally at T = 573 K. Solution of the riser energy balance assuming a typical value of the solids-to-wall heat-transfer coefficient of 150 W/(m2 K) indicates a maximum temperature variation of 3 "C. Contractor and Sleight (1987) listed isothermal operation as one of the inherent advantages of the CFB catalytic reactor for the production of maleic anhydride.

The reaction occurs in both the core and annular regions, the proportion of which is dependent on the local solid catalyst concentration as calculated by the hydrodynamic model. The simulation of the reaction requires solution of the mam balance for each species, i, reacted or formed. In the core region, where the gas is flowing, the steady-state mass balance includes a net convective input term plus the reaction term which is equal to the gas crossflow to the annulus. The reaction term ri in eqs 25 and 26, is positive for products formed and negative for reactants consumed. The core mass balance is written as follows (Patience, 1990):

In the annular region, there is no convective input term. The gas input to the annulus is due to crossflow from the core. The annular region mass balance is thus

2 k g r c ~ s ( 1 - cann)ri + (Ci ,c - Ci ,a) = 0 (26) (R2 - r:)

The units of the rate of reaction term, ri, in eqs 25 and 26 are mol/ (gcat.s). In order to incorporate the reaction rate expression into the mass balance equations, the pseudohomogeneous system was assumed. Such a reactor assumes no concentration gradients within a volume element and is valid for reactions which are kinetically controlled. The half-life of this reaction waa estimated assuming that the reaction occurs in a constant-volume batch reactor. This calculation results in a half-life of approximately 50

2656 Ind. Eng. Chem. Res., Vol. 31, No. 12, 1992

ms, indicative of an extremely fast reaction which would not normally be kinetically controlled. However, for a mean particle diameter of 75 fim, typical of the VPO catalyst used in the reactor, the Thiele modulus is of the order of and the effectiveness factor is 1. Thus, intraparticle resistance is negligible. Furthermore, the high slip velocities typical of circulating fluidized bed operation result in excellent mass transfer between the gas phase and the catalyst surface. Fbh (1986) indicated, in a qualitative manner, that the slip velocity is at a maximum in the circulating fluidized bed regime. Therefore, mass transfer or interparticle resistance will also be negligible. The negligible intraparticle and interparticle resistances indicate the reaction is kinetically controlled even though it is very fast, and thus the pseudohomogeneous assumption is valid for the reactor.

In this work, the simulation was run for a 20-m-high riser of 30-cm diameter. The riser is divided into 100 equal elements of 20 cm each. The length of the acceleration zone is calculated, and the hydrodynamic model is invoked to obtain the average axial voidage profile. From the Berruti-Kalogerakis model, the core voidage and core radius are obtained for each element. Forward finite dif- ferentiation is used to solve the core and annulus mass balances of eqs 25 and 26. For a given inlet concentration of reactants to the first volume element, the reaction rates may be calculated from eqs 3-5 using the kinetic data of Table 1. From the stoichiometry of the chemical reactions, the rates of water formation and oxygen depletion are calculated. It should be noted that since the majority of gas flows in the core region, the reactant concentration in the annular region in the first element is taken to be 0. This sets up the concentration gradient required for the mass transfer of gas from the core to the annulus.

Taking the gas velocity in the core (Vo in eq 25) as constant with height within each of the volume elements, then the axial concentration gradient of species i in the element dci/dx may be determined. If the partial deriv- ative is put in discrete form, and with the inlet concentrations to the first element known, the outlet concentration of species i from each element may be found. This concentration becomes the input to the next element where the mass balances of eqs 25 and 26 are performed again. These results then become the input to the next element, and the procedure is continued to the reactor exit,

By keeping track of the molar flows of each species reacted or formed, the moles of gas generated by the reaction will be known. This is converted to a volumetric flow of gas via the molar volume at reaction conditions. In this way, the increase in gas velocity in the riser due to gas generation may be calculated and the hydrodynamic model updated at the end of each interval, in order to evaluate the new flow structure corresponding to increasing values of the gas superficial velocity.

The n-butane conversion, yields of maleic anhydride and COP, and maleic anhydride selectivity are calculated at each 20-cm interval. The conversion, X, is defined as

(27) moles of butane converted

moles of butane fed X =

Maleic anhydride selectivity, S, is defined here as moles of maleic anhydride formed

moles of butane converted S = (28)

Finally, the product yield of MAN or COz, Y, is defined as

moles of product formed moles of butane fed Y = = SX (29)

5.0

5.0 mol% c g 20.0 mol% cB

3.5 - C, = 800 kg/m s Uoi = 6 m/s 0 m I. 3.0 -

T = 573 K

>

- 2 .5 -

b?

0 2 4 6 8 10 12 14 16 18 20

Riser Height. m

Figure 2. Single pasa butane conversion for various butane in air mixtures.

5. Simulation Results and Discussion 5.1. Effect of Feed Concentration. A mixture of

n-butane in air can form a potentially explosive mixture. The experiments of Wohlfahrt and Hoffman (1980) used a 1 mol % n-butane concentration while Sharma et al. (1991) limited their n-butane concentrations to less than 3 mol %. However, since the majority of the oxygen required for the reaction to proceed is supplied by the catalyst surface oxygen, and because catalyst regeneration occurs in a separate section, the CFB reactor may be operated in the absence of, or with small amounts of, gas-phase oxygen. As indicated by Contractor et al. (1988), this allows for operation with higher n-butane feed concentrations. The simulation was performed for feed concentrations of up to 50 mol % n-butane in air, and as indicated in Figure 2, conversion decreased with increasing n-butane feed concentration. Figure 2 also indicates that a limit is reached whereby an increase in the n-butane feed concentration has only a marginal effect on the conversion. Repeated runs of the simulation have indicated that for n-butane feed concentrations greater than 25 mol % , the riser conversion profile is essentially unchanged.

Figure 3 presents the rate of n-butane consumption to form both maleic anhydride and COP Although the n- butane concentration increases by a factor of 20, the reaction rate for 20 mol % n-butane is only 1.5 times the reaction rate at 1 mol % . The expected increase in reaction rate at higher n-butane feed concentrations is lessened by the inhibition term found in the denominator of eq 3. Conversion is defined in eq 27 as the moles of n-butane converted per mole of n-butane fed. For an increase in n-butane feed concentration, the rate of reaction does not increase proportionally, and thus conversion is decreased.

With the competing series and parallel reactions, the conversion includes formation of both desired and unde- sired products. Figures 4 and 5 compare the yields of maleic anhydride and COP along the riser length for two different &/butane mixtures. Figure 4 indicates that for 1 mol % n-butane in the feed, the yield of C02 is actually greater than the yield of maleic anhydride. When the feed concentration is increased to 5 mol % as shown in Figure 5, the yields of both maleic anhydride and COa decrease, but the yield of the desired product, maleic anhydride, is now greater than the yield of COO. This is obviously a more


: 0.020 m

0.010

0.100 I 1.0 mol% cB

G. = 800 kg/m2s UOi = 6 m / s

0.080 T = 573 K 20'o molX c B

e.

nv) 0.070 E L E 0.060 v

6 0.050 n 5 0.040 .- - YI c

0.030 /-

G, = 800 kg/m2s Uoi = 6 m/s

cg = 1.0 mol%

2.4 T = 573 K

0.0 I I 1 I I I I I I I 0 2 4 6 8 10 12 14 16 18 20

Riser Height. m

Figure 4. Single pass product yields for 1 mol % butane in air mixture.

profitable situation. These results indicate that although higher conversions are attained at relatively lower feed concentrations of n-butane, such an operating condition is not advantageous as the C02 yield exceeds that of the maleic anhydride. The ability to operate a CFB reactor in the absence of gas-phase oxygen now becomes important, as the resulting higher n-butane concentrations will result in improved yields of maleic anhydride.

Figure 6 shows the selectivity ,to maleic anhydride (MAN) along the riser defined in this work as the moles of MAN produced per mole of n-butane converted. Lower concentrations of n-butane in the feed result in lower selectivity. At concentrations of 20 mol % or greater, selectivities of 80% are realized. Experimental graphical data at 360 "C presented by Contractor et al. (1988) indicate selectivites of 75-80% for n-butane concentrations in the range of 12-50%. Thus the results of this work at 300 OC agree favorably with that data. The initial drop in selectivity seen a t the lower concentrations in Figure 6 may be attributed to the relatively greater initial oxygen

GI = 800 kg/m2 s Uoi = 6 m/s cg = 5.0 mol%

1.2 T = 573 K 1.4 t

; 0.8 c MAN Yield Y

.-

0.6

0.4

0.2 /

0.0 I I I I 4 1 I I I 0 2 4 6 8 10 12 14 16 18 20

Riser Height. m

Figure 5. Single pass product yields for 5 mol % butane in air mixture.

& 100.0

90.0 ......... ...................... .. ............... ... .. .................. .. ... ... ............. ,...., ,. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ - - -

_ _ - - - - ............. .........

80.0 .___.__- -- - - - - - - -

70.0 ' x - - - - _ - - _ _ _ _ _ _ _ _ _ _ - - - - - . i

2 bp 30.0

10.0

- 1.0 mol% cB - - - 5.0 mol% cB

20.0 mol% cg 50.0 mol% cg

.__--

..... ... .

G, = 800 kg/m2s Uo, = 6 m/s T = 573 K

n o 1 -.- 0 2 4 6 8 10 12 14 16 18 20

Riser Height, m

Figure 6. Maleic anhydride selectivity for various butane in air mixtures.

concentration in the feed. For a 1 mol % n-butane concentration in an air/butane mixture, the corresponding oxygen concentration is 20.8 mol %. Therefore, the initial reaction rate for COz formation is favored, and a drop in selectivity is observed. For 20 mol % n-butane in the feed, the initial oxygen concentration is 16.8 mol % and a flatter selectivity profile is seen. Once again, the advantage of operating the CFB reactor in the absence of gas-phase oxygen and increased n-butane concentrations is reinforced as the higher feed concentrations result in significantly improved selectivities. These selectivity results agree with the previous results of the product yields. Low yields at low n-butane concentrations correspond to low maleic anhydride selectivities.

The yield is defined in eq 29 as the product of the selectivity and the conversion. It has been shown in Figures 2 and 6 that, at high n-butane feed concentrations, conversion and selectivity profiles are constant. Therefore, the yield will also be constant. It was previously mentioned that approximately 25 mol % n-butane in the feed is the


0 . . .... ... .- r 3.0

0.40 1 I W 0.30 k$. I 0.20 \ '. \*..

0.00 1 ' I I I I I I I I 1 0 2 4 6 8 10 12 14 16 18 20

Riser Height, m

Figure 7. Effect of solids circulation rate on butane conversion and solids holdup.

critical concentration. Operation above this concentration will not significantly improve product yields or selectivity. Though the conversions and yields are very low, throughput from CFB reactors is very large and maleic anhydride production can be significantly higher than that from a fixed bed operation. Furthermore, there is the potential for enhanced reaction rates due to the cyclic operation between the riser and the regenerator (Matros, 1989; Lang et al., 1989). The cyclic operation has not been included in the simulation at this time due to lack of reaction kinetics for the process. Contractor et al. (1988) state that it is possible to improve this conversion by ad- dition of small amounts of gas phase oxygen in the riser. The potential for secondary air injection along the CFB riser would be valuable for this purpose.

5.2. Effect of CFB Operating Conditions on Reactor Performance. The use of the predictive hydrodynamic model in this simulation allows the effect of operating parameters such as solids circulation rate and superficial gas velocity to be carefully studied. These variables will affect the local solids catalyst holdup in the riser and hence the conversions and yields.

The n-butane conversion for solids circulation rates of 400, 600, and 800 kg/(m2.s) is presented along with the corresponding actual solids holdup for each circulation rate in Figure 7. The inlet gas superficial velocity is main- tained at 6 m/s, and n-butane concentration in the feed is 5 mol % . An increased solids circulation rate results in a greater solids holdup. This improves the contact between gas and solids, and therefore a higher conversion is achieved. It should be observed that, at the inlet of the riser, solids holdup is increased from about 28% solids at the low solids circulation rate to roughly 35% solids at the highest solids circulation rate. This increase in solids holdup is important because the catalyst will be moat active when it reenters the riser after regeneration in the standpipe.

In Figure 8, the n-butane conversion along the riser is presented for inlet superficial gas velocities of 4,5, and 6 m/s along with the corresponding solids holdups. The solids circulation rate and n-butane feed concentration are 800 kg/ (m2-s) and 5 mol % , respectively. The effect of increased superficial gas velocity is a dilution of the bed and thus a decrease in solids holdup. As a result, a de-

0.00 1 I I I I I 1 I I I 0 2 4 6 8 10 12 14 16 18 20

Riser Height. m Figure 8. Effect of inlet superficial gas velocity on butane conversion and solids holdup.

50% deactivation 70% deoctivotion

G, = 800 kg/m s Uo, = 6 m/s

I cg = 5.0 mol% 2.0 c C 0 .- E l T = 5 7 3 K

0.0 0 2 4 6 8 10 12 14 16 18 20

Riser Height, m

Figure 9. Effect of catalyst deactivation on single pass butane conversion.

crease in the conversion of n-butane is seen at higher gas velocities. Again, the dense bed at the base of the riser exhibita the greatest solids holdup.

Figures 7 and 8 confirm that it is beneficial to operate at the highest possible solids loadings and low gas velocities. The high initial reaction rates and the resulting gas generation in the dense bed will tend to increase the gas superficial velocity.

5.3. Results of Catalyst Deactivation Model. The preceding results were obtained assuming a catalyst deactivation corresponding to a value of 20% at the outlet of the riser. This section investigates the importance of catalyst deactivation as it is modeled here. Values of the decay constant a in eq 9 were chosen such that the catalyat was 20%) 50%, and 70% deactivated upon exiting to the cyclone. Feed concentration and riser operating conditions were held constant. As shown in Figure 9, increased catalyst deactivation reduces conversion along the riser. For approximately the first 3 m of riser length, the conversion is unaffected, but beyond this point the conversion


a rapid initial rise in reaction rate in the annular region. After 8-10 m, the rate of reaction levels off as reactants are depleted and mass transfer between the core and the annulus is lessened.

0.050 r 1

G, = 800 kg/m2s Uoi = 6 m/s

cg = 5.0 mol%

T = 573 K

0.045

E 0.040

c

I

0.000 I I I I I I I I I 0 2 4 6 8 10 12 14 16 18 20

Riser Height. m

Figure 10. Contribution to reaction by the core and annular regions for a 5 mol % butane in air mixture.

profiles begin to separate. At the riser inlet, the catalyst has just been regenerated and exposure to reaction conditions does not show a noticeable effect. Reaction occurs rapidly at this level, however, and after 3 m the trans- formation of the catalyst crystalline phases at the expense of the surface oxygen layers becomes significant. The higher degree of deactivation, resulting in reduced conversion, wil l consequently reduce the yields of both maleic anhydride and COP The results of this simple catalyst deactivation model indicate that reduction of the catalyst activity as the reaction proceeds has only a slight effect on the reactor performance. However, the results will underpredict the effect of deactivation on conversion due to the assumptions involved in the deactivation model. Further work to include the solids RTD in the simulation will produce more realistic results.

5.4. Influence of the Core and Annular Regions. It is interesting to investigate the separate contributions of the core and the annulus to the reaction. While the majority of the gas flows upward in the solids-lean core region, the annular region will have a higher concentration of catalyst. Furthermore, although the annulus is very dense, it is also very thin, which will affect the reaction rate per unit volume of bed.

Figure 10 separates the contribution to the reaction by the core and the annular regions. The model assumes that, initially, all of the gas fed to the riser is in the core region. The high initial concentration of reactants combined with the large solids holdup in the core at the base of the riser results in a high initial reaction rate in the core and essentially zero reaction rate in the annulus. The solids holdup in the core quickly decreases however, as illustrated earlier in Figures 7 and 8, and some of the upward flowing gas is transferred to the annulus. Also, the rapid initial reaction rate depletes a large portion of the reactants. These fadors combine to give the steep decline in reaction rate in the core seen in Figure 10. The gas concentration gradient between the core and the annulus is initially very large because all of the gas is concentrated in the core. Thus, with crossflow of gas modeled by a mass-transfer mechanism, the annulus gas concentration rises quickly. This fact, combined with the constant solids holdup corresponding to minimum fluidization voidage as assumed in the development of the hydrodynamic model, leads to

6. Conclusions The hydrodynamics of a circulating fluidized bed and

a fixed bed intrinsic kinetics model have been combined to form a comprehensive computer simulation of the partial oxidation of n-butane to maleic anhydride. The model allows investigation of the effect of changes in operating conditions or reactor geometry, important in reactor design and process control. Key results presented indicate decreased conversion with increased n-butane feed concentration, but improved selectivity at the higher n- butane concentrations. For high solids circulation rates, conversion is increased due to increased solids holdup, and thus better gasaolids contacting. Increasing the gas superficial velocity decreased conversion due to the dilution of the bed or decreased solids holdup. Catalyst deactivation had a only a slight effect on reactor performance, however; more work is needed to make the deactivation model more rigorous.

Acknowledgment

The work reported in this paper has been supported by a Natural Sciences and Engineering Research Council of Canada operating grant and scholarship.

Nomenclature a = deactivation parameter (8-l) cB = n-butane concentration (mol/L) c , = oxygen concentration (mol/L) cMAN = maleic anhydride concentration (mol/L) q C = concentration of the ith species in the core (mol/L) ci,, = concentration of the ith species in the annulus (mol/L) Dm = molecular diffusivity of n-butane in air (m2/s) D, = mean particle diameter (m) D, = riser diameter (m) fi0 = smooth pipe interfacial friction factor g = gravitational acceleration constant (m/s2) G, = overall solids circulation rate (kg/(m2 8 ) ) k = proportionality constant in the acceleration zone (m-l) kl = kinetic constant for maleic anhydride formation (mol'-*

k z = kinetic constant for COz formation (moll-8 LB/(g a)) k3 = kinetic constant for MAN decomposition (mol" L1+/(g

kd = kinetic constant for catalyst surface deactivation (s-l) k, = gas mass-transfer coefficient from core to annulus (m/s) K,, = annulus-to-core solids interchange coefficient (m/s) K,, = core-to-annulus solids interchange coefficient (m/s) KB = equilibrium constant in Centi kinetics (mol/L) Lac, = length of the acceleration zone (m) P = total pressure (Pa) qo = number of deactivated catalytic sites for a totally

deactivated catalyst q ( t ) = number of deactivated catalytic sites at any time t rl, r m = rate of maleic anhydride formation from n-butane

(mol/(g 9)) r2, rCOl = rate of C02 formation from n-butane (mol/(g a)) r,, -rmN = rate of maleic anhydride decomposition (mol/(g

8 ) ) r, = rate of the ith reaction, where i = 1, 2, 3 (mol/(g 8)) r, = core radius (m) rd = rate of catalyst deactivation (mol/(g 8 ) ) R = riser radius (m) Re = Reynolds number = U o p p p / p Re, = particle Reynolds number = fJ$gDp/bg

L"/(g 5 ) )

8))


S = product selectivity, moles of product formed per moles

S = Stokes number = pJl,2Uo/18pgUt t , = solids residence time (s) T = temperature (K) U,,, = combined phase core velocity (m/s) Uo = riser superficial gas velocity (m/s) Ua = average solids velocity (m/s) U,, = core solids velocity (m/s) U,, = solids velocity at the end of the acceleration zone (m/s) Ut = terminal settling velocity of a single solids particle (m/s) x = axial location in the riser (m) X = n-butane conversion, moles converted per moles fed Y = product yield, moles of product formed per moles fed Greek Letters a, 8, 6 , = kinetic constants, exponents in Centi rate ex-

pressions r = constant in acceleration zone tan = annular voidage fapp = apparent axial voidage in the acceleration zone taVg = average axial voidage t b = apparent voidage at riser bottom tC = core voidage emf = voidage at minimum fluidization conditions to = actual voidage at riser bottom t, = voidage at the end of the acceleration zone { = constant in K,, expression pb = bulk density of solids (kg/m3) pcom = combined phase core density (kg/m3) pg = gas density (kg/m3) pa = solid particle density (kg/m3) cp = slip factor 4(t) = fraction of the total number of catalytic sites deacti-

vated Registry No. Butane, 106-97-8; maleic anhydride, 108-31-6;

converted

vanadium phosphorus oxide, 65506-75-4.

Literature Cited Arena, U.; Cammarota, A.; Pistone, L. High Velocity Fluidization

Behavior of Solids in a Laboratory Scale Circulating Bed. In Circulating Fluidized Bed Technology; Basu, P., Ed.; Pergamon Press: Toronto, Ontario, 1986; Vol. 1, Chapter 2.

Bader, R.; Findlay, J.; Knowlton, T. M. Gas/Solids Flow Patterns in a 30.5 cm Diameter Circulating Fluidized Bed. In Circulating Fluidized Bed Technology ZI; Basu, P., Large, J. F., Eds.; Perga- mon Press: Toronto, Ontario, 1988; Vol. 1, Chapter 3.

Berruti, F.; Kalogerakis, N. Modelling the Internal Flow Structure of Circulating Fluidized Beds. Can. J. Chem. Eng. 1989, 67, 1010-1014.

Bolton, L. W.; Davidson, J. F. Recirculation of Particles in Fast Fluidized Risers. In Circulating Fluidized Bed Technology ZI; Basu, P., Large, J. F., Eds.; Pergamon Press: Toronto, Ontario, 1988; Vol. 1, Chapter 3.

Centi, G.; Fornasari, G.; Trifirb, F. n-Butane Oxidation to Maleic Anhydride on Vanadium-Phosphorus Oxides: Kinetic Analysis with a Tubular Flow Stacked-Pellet Reactor. Znd. Eng. Chem. Prod. Res. Dev. 1985,24,32-37.

Centi, G.; TrifirB, F.; Ebner, J. R.; Franchetti, V. M. Mechanistic Aspects of Maleic Anhydride Synthesis from C, Hydrocarbons over Vanadium Phosphorus Oxides. Chem. Rev. 1988,88,55-80.

Contractor, R. M. Butane to Maleic Anhydride in a Recirculating Solids Reactor. In Circulating Fluidized Bed Technology ZI; Basu, P., Large, J. F., Eds.; Pergamon Press: Toronto, Ontario, 1988; Vol. 1, Chapter 8.

Contractor, R. M.; Sleight, A. W. Maleic Anhydride from C, Feed- stocks Using Fluidized Bed Reactors. Catal. Today 1987, 1, 587-607.

Contractor, R. M.; Chaouki, J. Circulating Fluidized Bed as a Cata- lytic Reactor. In Circulating Fluidized Bed Technology ZZI; Basu, P., Horio, M., Hasatani, M., Eds.; Pergamon Press: Toronto, Ontario, 1990, Vol. 1, Chapter 1.

Contractor, R. M.; Bergna, H. E.; Horowitz, H. S.; Blackatone, C. M.; Chowdhury, U.; Sleight, A. W. Butane Oxidation to Maleic An- hydride in a Recirculating Solids Reactor. In Catalysis 1987; Ward, J. W., Ed.; Elsevier: Amsterdam, 1988; Vol. 38, Chapter 4.

DeMaio, D. A. Will Butane Replace Benzene as a Feedstock for Maleic Anhydride. Chem. Eng. 1980,87 (lo), 104-106.

Enick, R.; Falkenberg, K. L.; Klinzing, G. E. Acceleration Length Model for Pneumatic Transport. In Particulate and Multiphuse Processes; Ariman, T., Vezirglu, T. N., Eds.; Hemisphere: New York, 1987; Vol. 1, Chapter 6.

Higbie, R. The Rate of Absorption of a Pure Gas Into a Still Liquid During Short Periods of Exposure. Trans. AZChE. 1935,31 (2), 365-389.

Irving-Monshaw, S.; Kislin, A. Lively Marketa Brighten Maleic An- hydride Horizons. Chem. Eng. 1989, 94 (ll), 29-33.

Lang, X.; Hudgins, R. R.; Silveston, P. L. Application of Periodic Operation to Maleic Anhydride Production. Can. J. Chem. Eng. 1989,67,635-645.

Matros, Y. s. In Catalytic Processes Under Unsteady-State Con- ditions; Elsevier: Amsterdam, 1989; Vol. 1, Chapter 1.

Patience, G. S. Circulating Fluidized Beds; Hydrodynamics and Reactor Modelling. PbD. Dissertation, Ecole Polytechnique de Montriial, MontrBal, PQ, 1990.

Patience, G. S.; Chaouki, J. Catalytic Oxidation of n-Butane to Maleic Anhydride in a Fast Fluidized Bed Reactor. Presented at the CHISA Tenth International Conference on Chemical Engi- neering, Prague, 1990; paper 556.

Patience, G. S.; Chaouki, J.; Berruti, F.; Wong, R. Scaling Consid- erations for Circulating Fluidized Bed Catalytic Risers. Powder Technol. 1993, in press.

Reh, L. The Circulating Fluidized Bed Reactor-A Key to Efficient Gae/Solid Processing. In Circulating Fluidized Bed Technology; Basu, P., Ed.; Pergamon Press: Toronto, Ontario, 1986; Vol. 1, Chapter 1.

Senior, R. C.; Brereton, C. Modelling of Circulating Fluidized Bed Flow Structure and Heat Transfer; Internal Report by the De- partment of Chemical Engineering, University of British Colum- bia, 1990.

Sharma, R. K.; Cresswell, D. L.; Newson, E. J. Kinetics and Fixed Bed Reactor Modelling of Butane Oxidation to Maleic Anhydride. MChE J. 1991, 37 (l), 39-47.

Smith, J. R. Kinetics of Catalyst Deactivation. In Chemical Engi- neering Kinetics, 3rd ed.; McGraw-Hill: New York, 1981; pp 381-383.

Takeuchi, H.; Hirama, T. Flow Visualization in the Riser of a Cir- culating Fluidized Bed. In Circulating Fluidized Bed Technology IZfi Basu, P., Horio, M., Haaatani, M., Eds.; Pergamon Press: Toronto, Ontario, 1990; Vol. 1, Chapter 2.

Weinstein, H.; Shao, M.; Schnitzlein, M. Radial Variation in Solid Density in High Velocity Fluidization. In Circulating Fluidized Bed Technology; Basu, P., Ed.; Pergamon Press: Toronto, On- tario, 1986; Vol. 1, Chapter 2.

Wohlfahrt, K.; Hoffmann, H. Kinetics of the Synthesis of Maleic Anhydride from n-Butane. Chem. Zng. Tech. 1980,52,811-814 (in German).

Wong, R.; Pugsley, T.; Berruti, F. Modelling the Axial Voidage Profile and Flow Structure in the Riser of Circulating Fluidized Beds. Chem. Eng. Sci. 1992,47 (9-11), 2301-2306.

Received for review March 23, 1992 Revised manuscript received August 10, 1992

Accepted September 4,1992

Documents

maleic anhyride production method