Ind. Eng. Chem. Res. 1994,33, 3053-3062
3063
Fluid Catalytic Cracking Catalyst for Reformulated Gasolines. Kinetic Modeling? Arturo Gianetto,"Hany I. Farag? Albert0 P. Blasetti,"and Hugo I. de Lasa' Chemical Reactor Engineering Centre, Faculty of Engineering Science, University of Western Ontario, London, Ontario N6A 5B9,Canada
Changes of the relative importance of intradiffision on USY zeolite crystals were studied as a way of affecting selectivity of catalytic cracking reactions. Zeolite crystals were synthesized ( S U N = 2.41, activated and stabilized using ion exchange and steam calcination to obtain USSY (Ultra Stable Submicron Y) zeolites. After the activation the zeolites were pelletized (45-60 pm particles). USSYs were tested in a novel Riser Simulator. Results obtained show that total aromatics (BTX),benzene, Cq olefins, and coke were significantly affected with the change of zeolite crystal sizes. Gasolines produced with USSY zeolites contain less aromatics and particularly lower benzene levels. Experimental results were analyzed with a model including several lumps: unconverted gas oil, gasoline, light gases, and coke. This model also accounts for catalyst deactivation as a function of coke on catalyst. Various kinetic parameters were determined with their corresponding spans for the 95% level of confidence.
1. Introduction Over the past two decades, considerable attention has been devoted to overcome air pollution problems associated with the emissions of gasoline-fueled vehicles. A prime cause of air pollution is the emission of partially combusted products (carbon monoxide, oxides of nitrogen), unburnt hydrocarbons, and fuel from gasoline tanks and the fuel-delivery system (Seddon, 1992). Pursuant t o new legislation under the United States Clean Air Act of Nov 15, 1990, the production of gasolines will require significant changes, such as a decrease in aromatics, in order to overcome pollution associated with the emissions of gasoline-fueled vehicles. Concerning benzene, an aromatic of well-known carcinogenic effects, it was legislated to be limited to 1% by volume, with the total aromatics capped a t 25% by volume. Production of reformulated gasoline affects several unit operations in a conventional oil refinery (Stokes et al., 1990). Because fluid catalytic cracking (FCC) currently produces less aromatics than reforming, FCC catalysts will play a greater role in the production of reformulated gasoline. FCC converts gas oil and other high-boilingdistillable products into gasoline-range hydrocarbons. Olefin content is very much dependent on catalyst type, reactor temperature, and catalystloil ratio. Thus, octane numbers can be increased by increasing olefin content. Higher amounts of olefins can be handled by further downstream processing (e.g., alkylation). For reformulated gasoline, it would be better if some of the aromatics were replaced by isoparaffins. ~~
+ Presented at the Symposium on Catalytic Reaction Engi-
neering for Environmentally Benign Processes, National Meeting of the American Chemical Society, San Diego, March 1318, 1994. * Author to whom correspondence should be addressed. E-mail:
[email protected]. EniChetn, Baytown Plant, 4803 Decker Drive, Baytown, TX 77521. 5 Department of Chemical Engineering,Norwegian Institute of Technology, Trondheim, Norway. Departamento de Procesos, Facultad de Ingenieria, Universidad de La Patagonia, Comodoro Rivadavia, Chubut, Argentina.
*
0888-5885I94l2633-3053$04.50/0
The objective of this study is to determine t o what extent intraparticle diffusion may affect the catalytic cracking process. In particular, it is important to determine whether the crystal size of the Y zeolites, which are the active phase of FCC catalysts, can affect gasoline composition (Rajagopalan et al., 1986). Another objective of the research is to determine whether it is possible to decrease the amount of benzene and BTX aromatics in gasoline. Once the zeolites with different crystal sizes but similar chemical composition are synthesized, it was planned to pelletize these zeolites by spray drying getting pellets in the 60 pm range. Using the Riser Simulator, kinetic constants for the two catalysts were determined using a four-lump model highlighting light gases, coke, gasoline, and unconverted feedstock (Farag et al., 1993). Given the ample range of operating conditions studied, including changes in temperature, catalystloil, and the reaction time, it was expected that the kinetic model obtained could provide information concerning other relevant kinetic parameters such as energy of activation and changes of energy of activation with zeolite particle size.
2. Experimental Section 2.1. Catalyst Preparation. Zeolites prepared in this study, as described by Gianetto (19931,had the same chemical properties but different crystal sizes. Zeolites were activated and hydrothermally stabilized to obtain USSY zeolites through a process of ion exchange and steaming. They were finally pelletized using spray drying. The catalyst was steamed in a minifluidized bed capable of holding up to 20 g of zeolites and pellets. The steaming unit was designed to operate at atmospheric pressure and at a temperature ranging between 600 and 800 "C. The catalyst was held in place by a porous disk made of inconel which has a pore diameter of about 20 pm. Spray-drying techniques were employed for pelletization of the zeolites producing pellets with sizes and zeolite loadings matching, as closely as possible, those used in commercial FCC risers: 50-60 pm and 23 wt %. T w o streams of hot air were fed t o the spray drier: a flow rate of 0.3 m3/min at 270 "C; a secondary side flow of 0.1 m3/min at 245 "C. The secondary stream of 0 1994 American Chemical Society
3054 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 Table 1. Effect of Steaming on the Unit Cell Size of NaY Zeolites: Unit Cell Size for the So-called Small and Large Zeolites large (A) small (A) after synthesis 24.68 24.67 first steaming 24.41 24.39 second steaming 24.32 24.29 Table 2. Calculated Crystallinity of Large and Small Zeolites synthesized samdes 9% XRD large 35.7 small 28.5
hot air was fed tangentially to the drying chamber providing longer particle residence time and increased yields of pellets (Gianetto, 1993). 2.2. Catalyst Characterization. Catalyst synthesized according to the methods described above was analyzed to determine physical and chemical characteristics. X-ray diffraction (XRD)was used for identification of the crystal structure: unit cell size, crystallinity, and crystal size. Unit cell size being an indicator of crystal framework composition allowed checking of the SYAl crystal ratio before and after steaming (dealuminization). Unit cell size (Table 1)was calculated for both “USSYsmall” and “USSY-large”zeolites subject to the following conditions: (1)after synthesis; (2) after the first steaming-calcination treatment at 770 “C; (3) after the second steaming-calcination at 765 “C. It is interesting to note that the unit cell size decreased consistently for each of the two crystals and this was apparent after the first steaming-calcination. As reported in Table 1,unit cell size evolved from 24.67-24.68 A (after synthesis) to 24.29-24.32 A for USSYs (after the second steamcalcination). These values are typical of the ultrastable Y zeolites, and it demonstrates that dealumination in USSY crystal framework was quite effective with alumina being replaced in the framework by silica. Crystallinity was assessed comparing prepared zeolites with a standard sample, with assumed 100% crystallinity equivalent (ASTM D 3906-85). A standard NaY sample, from Stream Chemicals, was treated following methods and procedures described for the calculation of the unit cell size. The results of the above calculations for both USSY-large and USSY-small are given in Table 2, showing zeolites after steaming calcination, kept about 28.5%-35.7% crystallinity. Crystal sizes were evaluated using the relationship between crystal size and X-ray line broadening introduced by Scherrer in 1918 as reported by Leviosiforich (1964). Following this method the average crystallite dimension, D, is considered directly proportional to the wavelength, I in angstroms, and inversely proportional to the pure diffraction broadening, p, and the cosine of the Bragg angle, 8.
The constant K depends largely upon the crystallite shape with values ranging between 0.7 and 1.7. If is taken as the half-maximum line breadth, the value of K is 0.9. It has to be stressed that no broadening appears to take place in the XRD patterns for 1 and 10 pm crystals, and thus, the method could be used for determination of submicron zeolite crystals only. Quartz used as an indicator together with the zeolite samples,
Table 3. Mean Crystal Sizes of the T w o Zeolite Samples, Large and Small size large small
mean crystal size D(hk0 @m) 0.413 0.124
Table 4. Effect of Steam Calcination and Ion Exchange on the Chemical Structure of NaY Zeolite large large small small treatment SiiAl N a % SVAl N a % synthesized 2.35 10.85 2.44 9.27 first exchange 2.39 1.78 2.47 1.45 first steaming and third exchange 2.38 1.19 2.43 1.27 second steaming 2.43 1.20 2.43 1.17 Table 5. Changes in the Surface Area of the Zeolite Crystals from the Synthesized Form to the After Treatment Form small 2 large 2 treatment (m2/g) (m2/g) synthesized 523 525 second steaming 263 269
USSY-small and USSY-large, had crystal sizes greater than 5 pm. Therefore, the width of the resulting peak located at the d-spacing value of 3.34 was solely due to the XRD. The value of p was determined by first finding the ratio blB‘, where b is the width of the quartz peak and B’ is the width of the peak of the zeolite sample analyzed. Using the functional relationship, as described by Leviosiforich (19641, it is possible to determine the ratio of BIB‘ and therefore the value of p. Equation 1was applied to determine mean dimensions of the crystallite perpendicular to the diffracting planes (hkl). As reported in Table 3 crystal sizes obtained for the USSY-small zeolites were in the 0.12 pm range while those for the USSY-large were in the 0.4 pm range. The atomic composition of each of these two samples was determined by using SEM-EDX. The Na in the crystal structure of the NaY zeolite (Table 4) showed changes during the steam-calcination and ion exchange processes. It can be noticed, however, that the SifAl ratio does not vary after the first and second steaming calcination, with amorphous AI remaining presumably locked in the pores of the catalyst. Moreover, the ion exchange process used to extract the sodium from the crystal structure was successful, as reported in Table 4, since the Na decreased from 10.8%to 1.2% and from 9.27% to 1.17% for USSY-large and USSY-small, respectively. Dealumination of zeolites translates into removal of the alumina from the framework. Microprobe studies have shown that alumina migrates to the outside surface of the zeolite particle during the dealumination process, probably through the mesopore system (Peters, 1992). As a result, nonframework alumina tends to concentrate in the outer region of particles with overall SUAI remaining unchanged (Jacquinot et al., 1990). Thus, major changes occur in dealuminized zeolites with formation of nonframework alumina, formation of mesoporosity, and reduction of aluminum acid sites in the framework (Peters, 1992). Specific surface area of the crystal structure was determined with a TPD/TPR 2900 Analyzer from Micromeritics. The BET surface area analyses (Table 5) permitted evaluation of the catalyst surface area (zeolite crystals). The surface area of the crystalline phase for
Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 3055 Table 6. Temperature of Desorption and TCD Area per gram of FCC Catalyst for the Weak and Strong Acid Sites of Large and Small Zeolites weak acid sites temperature ("C) TCD aredg of FCC temperature ("C) TCD aredg of FCC
(a) Large Zeolites 158 5.0089 x lo6 (b)Small Zeolites 157 5.6 x lo6
strong acid sites 695 10.78 x 726 11.7 x
lo6
lo6
the two zeolites synthesized was determined to be 525 and 523 m2/g for USSY-large and USSY-small, respectively. The two zeolite crystals were tested again after the second steaming-calcination. The surface area for USSY-large was found to be 269 m2/g,while for USSYsmall it was found to be 263 m2/g. This further confirms that some of the zeolite crystals were transformed into amorphous material during the second steamingcalcination treatment. These results were also consistent with results obtained with the XRD pattern of the same samples in which the base line of the XRD patterns was shifting upward indicating increase of the amorphous phase. Results obtained from the TPD analyses suggest that two different acid sites are present in the USSY, as observed through ammonia desorption: stronger and weaker acid sites (Table 6). Thus, the stronger and weaker acid sites appear to share acidity in the zeolites synthesized with the USSY-small showing higher acid strength for the stronger sites: the ammonia desorption temperature for USSY-small was 726 "C instead of 695 "C for the USSY-large. 3. Kinetic Modeling of Catalytic Cracking
Lumping techniques have been used to develop kinetic models for catalytic cracking (Wei and Kuo, 1969; Weekman, 1979). Lumping techniques are useful given the large number of individual species present in the gas oil feedstock with boiling point ranging between 220 and 530 "C. Due to the large number of chemical species, one possible approach is to group them into a smaller group of pseudospecies to obtain a tractable number of kinetic equations. Species can be lumped together only if the dynamic behavior of the resulting pseudospecies is independent of the species composition (Coxson and Bischoff, 1987). A three-lump model was developed based on the unconverted feedstock, gasoline, and light gases plus coke (Weekman and Nace, 1970). More sophisticated lumping models were developed in which the feedstock could be further divided to account for molecular type (Jacob et al., 1976). Recent studies consider the need of coke separation from the light gases, creating a fourlump model as follows: unconverted feedstock, gasoline, light gases, and coke (Farag, 1993). There are many types of catalytic reactors available to determine the kinetic rate parameters, such as fixed bed, fluidized bed, stirred batch, continuous stirred tank, recirculating transport differential, straightthrough transport, and pulse microreactor. In the past, researchers relied mainly on the use of fured bed to study catalytic cracking reactions or the so-called MAT (Micro-ActivityTest). Although these reactors are presently used for catalyst testing, they pose certain important limitations for kinetic modeling (Farag, 1993). First, kinetic data are obtained using catalyst time on stream of about 1-2 min, which does not represent the
catalyst contact time of the commercial riser. Second, kinetic modeling in these reactors involves a cumbersome mathematical process given temperature gradients, due to the endothermicity of the reaction, and coke profiles. Continuous fluidized beds also present problems for kinetic modeling given the complex interaction between kinetics and hydrodynamics (Kraemer, 1991). To obtain reliable kinetic data for riser catalytic cracking, equipment must be used that allows for short contact times and to mimic conditions that closely simulate those found in the riser. The Riser Simulator, designed and developed at the Chemical Reactor Centre at the University of Western Ontario (de Lasa, 1991; Kraemer, 19871, provides an alternate and most effective reactor for studying catalytic cracking. Additional information about the performance of the Riser Simulator unit for kinetic modeling can be found in a recent review by de Lasa and Kraemer (1992). The Riser Simulator simulates catalytic cracking reactions under conditions similar to those taking place in a commercial risers. It provides short reaction times and behaves like a batch fluidized bed, avoiding coke profiles and gas channeling. The Riser Simulator, with its high gas phase recirculation rates, provides intense gas-phase mixing approaching ideality and minimizing concentration gradients (Pekediz et a1.,1992). The Riser Simulator also provides high volumetric recirculation rates in the context of a well-mixed gas phase continuously reacting with the catalyst. Thus, the catalyst particles are over time in contact in the Riser Simulator with a changing hydrocarbon environment simulating the catalystreactant kinetics rates of a commercial riser. In the Riser Simulator, a fluidized catalyst bed is created. When the shaft is rotating, the gas is forced outward from the center of the impeller, thus creating a lower pressure in the center region of the impeller which results in an upward flow of gases from the bottom of the reactor annular region where the pressure is slightly higher. Since the Riser Simulator provides intense mixing of gas and solids, it limits the formation of coke profiles and gas channeling that occur in a fixed bed of fine solids, such as FCC catalysts with an average particle size of 60 pm. Therefore, in the Riser Simulator the catalyst sample can be considered homogeneous in activity at specific contact times. To obtain reliable data from the Riser Simulator, it is important to use operating conditions found in commercial units: oil partial pressure ranging between 0.8 and 2 atm, catalyst-to-oil ratio ranging between 3 and 10, temperature ranging between 500 and 550 "C, and reaction time less than 10 s. The Riser Simulator operates with a fixed amount of gas oil fed with an automatic injector and vaporized almost instantaneously. Following this and after a preset reaction time, the products are quickly evacuated and diverted toward a computer-controlled sampling valve and gas chromatographic (GC) system (Gianetto, 1993). A diagram describing the Riser Simulator and the associated sampling system is presented in Figure 1. Given the Riser Simulator, a constant volume reactor, operates in the batch mode the change in the total number of moles of gas oil over change in reaction time and total reactor volume is considered equal to the cracking rate of gas oil. Therefore, balances in the Riser
3056 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 ARGONIAIR
-dCA/dt = 4(k1
+ k,)CpmflT
(7)
Moreover, the gas oil concentration in the reactor can be evaluated as follows:
C A = YAwd(J(MAVT)
where YAis the weight fraction of gas oil, W T is the total weight of hydrocarbon species in the reactor, and M A is the average molecular weight of gas oil. Once CAis substituted into the gas oil mass balance equation, this relationship can be expressed in terms of weight fractions:
TO GC
Y
(8)
VENT/ VACUUM
-dYA/dt = 4(kl
GLASS CHAMBER
Figure 1. Schematic description of the Riser Simulator and the sampling system including the heated vacuum box, the four-port valve, the six-port valve, and the glass chamber.
+ ~ ~ ) Y ~ W T ~ J ( M A V , "(9))
Regarding the rate of gasoline formation, it contains two terms, one for the formation of gasoline result of gas oil cracking and the other for the disappearance of gasoline due to gasoline cracking:
Simulator can be expressed as follows: rB
-dCA/dt = rAmflT
(2)
where r A is the rate of gas oil cracking expressed in moles of oil per gram of catalyst per second, CAis the concentration of gas oil expressed in moles of gas oil per unit of reactor gas volume, mc is the mass of catalyst, and VT is the total reactor gas volume. The four-lump model can be adopted for simulation of FCC units and will be the basis of modeling in the present study with the various kinetic constants
A (gas oil)
(10)
Thus, the change in gasoline concentration (CB)in the Riser Simulator, with the assumption that all sites leading to either gasoline, light gases, or coke are deactivated with the same function,
4 = 41 = 42
(11)
can then be written as dCB/dt = 4[ulklC2 - k2C~]mc/V,
(12)
where u1 is the stoichiometric coefficient for gasoline formation equal to (MAIMB)ratio between molecular weight of gas oil and molecular weight of gasoline respectively. Furthermore, gasoline concentrations can be expressed in terms of weight fractions as follows:
-
B (gasoline) k22
D (coke)
defined as follows:
k, = k,
+ k3
(3)
= '31
+ '32
(4)
3 '
= 4u1k1c2 - 4lk2CB
dYB/dt = f$[Wd(J(MAVT)k1Y2 - k2YB]mflT (13) Similarly, the mass balance for the light gases and for the coke can be written as:
+
dYcldt = @ [ ( W ~ M A V T ) ~k ~ 2 1~YY~~l m f l ~ (14) This model involves four simultaneous equations describing the evolution of unconverted gas oil lump (A), the gasoline lump (B), the light gases lump (D), and the coke lump (C). The cracking of gas oil, rA, can be described using a second order reaction rate (de Lasa and Kraemer, 1992). The following considerations are adopted: cracking rates are functions of reactant partial pressures; reactant partial pressures can be expressed, using the ideal gas law, as functions of chemical species concentrations. rA= -4(kl
+ k3)C2
(6)
where 6 represents the catalyst activity decay due to coke deposition and ( K I + K3) is the overall gas cracking rate constant, the sum of individual constants kl and k3 from the four-lump scheme. By substituting the value of r A into the gas oil differential balance in the Riser Simulator, the following relationship results:
and dYddt = @ [ W ~ ( J ( M A V T4-) k~2~2~YY~~l m f l ~ (15) Furthermore, if an activity decay function based on coke-on-catalystconcentration is postulated, the following differential relationship can be considered to express the changes of the unconverted gas oil lump: -dCA/dt = k, exp[-aYDw~m,]CA2m,jVT (16) Proceeding with similar transformations, as the ones considered above relating CA and YA, the following relationship is obtained: -dYA/dt = k, ~ ~ ~ [ - ~ Y D ~ ~ ~ , ] Y A ~ W T (17) ~J(MAVTP In summary, model equations are available to follow the cracking of gas oil in the Riser Simulator. A fourlump model, highlighting unconverted gas oil, gasoline,
Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 3057 Table 7. Gas Oil fioperties specific gravity aniline point ("C) Conradson carbon (wt %) Ni (ppm) V (ppm) volumetric average bp ("C) simulated distillation ("C) IBP 5wt% 10 wt % 30wt% 50 w t % 70wt% 90 wt % 95 wt %
FBP paraffins (wt %) molecular weight naphthalenes (wt %) molecular weight aromatics (wt %) molecular weight average molecular weight
0.9389 59 0.15 0.006 i0.005 393 261 300 322 357 387 423 474 498 544 7.4 313 40.9 305 51.7 358 333
Table 8. Conversions and Selectivities for Small, c/o = 4.7"
550 525 500
4. Conversions and Selectivities
Runs at three residence times (3, 5, and 10 s), three temperatures (500, 525, and 550 "C), and two catalystto-oil ratios (C/O = 3.6 and 4.79) were developed during the present study using a synthetic gas oil feedstock (Table 7). Runs were repeated when needed to obtain adequate confidence minimizing errors. The USSY-large and USSY-small catalysts were extensively tested showing these zeolites were thermally stable and very active. Thermal stability of these two catalysts was proven given catalyst activity did not change after prolonged exposure of the catalyst at high temperatures (500-550 "C) with catalyst regenerations a t 650 "C. Furthermore, moderate higher catalyst activity was obtained for these two catalysts when compared with GX-30 and Octacat commercial catalysts tested in the same Riser Simulator (Kraemer, 1991). This higher activity can be explained considering both USSY-large and USSY-small had zeolites with mean crystal sizes (0.413 and 0.123 pm) smaller than currently used commercial catalysts (about 1pm). Therefore, in the USSYs the anticipated result of smaller diffusional controls or higher effectiveness factors led to increased gas oil cracking rates. Regarding the conversion of gas oil, it was determined using a weight basis. Thus, the calculated values were obtained from the chromatographic report of the GC integrator. This was done by assuming the unconverted gas oil to consist of compounds with a carbon number greater than or equal to C15. Small corrections were needed, however, in the GC gas oil conversion to incorporate coke yields. Gas oil conversions for both USSY-small and USSYlarge, are listed in Tables 8-11. The general trend of the gas oil conversions for the two catalysts shows that the conversions tend to increase as the temperature and residence time increase. Since both catalysts, USSYlarge and USSY-small, were very active, relatively small changes in the amount of feedstock injected (catalyst/
18.74 13.29 19.7 17.64 20.06 27.03 23.56 23.64 28.74
26.5 32.66 28.72 27.39 25.9 24.26 26.56 22.55 17.01
52.26 52.13 50.07 52.92 52.32 47.7 48.06 52.35 51.94
2.5 1.91 1.5 2 1.7 1
1.8 1.48 1.29
81.26 86.71 80.3 82.36 79.94 72.97 76.44 76.36 71.26
0.643 0.0307 0.601 0.0220 0.624 0.0186 0.643 0.0242 0.654 0.0212 0.654 0.0137 0.629 0.0235 0.686 0.0193 0.729 0.0181
GO, unconverted gas oil; LG, light gases; GA, gasoline; sel, selectivity.
cS-c12
Table 9, Conversions and Selectivities for Small, C/O = 3.8"
550 525
light gases, and coke, is postulated. This model is applicable given the conditions of intense mixing for both gas phase and catalyst prevalent in the Riser Simulator (Pekediz, 1994).
10 5 3 10 5 3 10 5 3
500
(I
10 5 3 10 5 3 10 5 3
19.03 19.17 18.33 20.05 24.78 27.87 27.99 23.83 26.81
24.05 27.37 28.16 20.1 19.95 20.57 16.12 18.53 19.61
55.46 52.22 52.59 58.61 53.54 50.84 54.68 57.04 53.07
1.4 1.18 0.9 1.2 1.7 0.7 1 0.7 0.5
80.97 80.83 81.67 79.95 75.22 72.13 72.01 76.17 73.19
0.685 0.646 0.644 0.733 0.712 0.705 0.759 0.749 0.725
0.0172 0.0145 0.0110 0.0150 0.0226 0.0097 0.0138 0.0091 0.0068
GO, unconverted gas oil; LG, light gases; GA, gasoline; sel, selectivity.
cS-c12
Table 10. Conversions and Selectivities for Large,
c/o = 4.7a
temp time ("C) (s) 550 525 500
10 5 3 10 5 3 10 5 3
GO
LG
GA
coke conv
(%)
(%)
(%)
(%)
(%)
GA sel
cokesel
14.2 20.37 24.43 18.26 24.42 24.34 24.64 25.25 31.88
34.27 28.73 27.97 26.08 26.06 29.3 24.34 25.87 23.32
46.2 47.14 45.08 51.63 46.53 44.14 47.91 46.64 42.9
5.32 3.74 2.5 4 2.97 2.2 3 2.2 1.82
85.8 79.63 75.57 81.74 75.58 75.66 75.36 74.77 68.12
0.538 0.592 0.597 0.632 0.616 0.583 0.636 0.624 0.630
0.0620 0.0469 0.0330 0.0489 0.0392 0.0290 0.0398 0.0294 0.0267
GO, unconverted gas oil; LG, light gases; GA, gasoline; sel, CS-c12 selectivity.
Table 11. Conversions and Selectivities for Large, C/O = 3.6" temp time GO LG GA coke conv GA ("C) (s) (%I (%) (%) (%I (%) sel coke sel 550 525
10 5 3 10 5
500
a
3 10 5 3
22.21 23.99 25.74 28.21 23.95 23.07 29.89 20.36 26.62
28.53 26.3 28.58 23.72 26.53 29.13 22.27 30.29 33.26
47.01 .48.07 43.88 45.99 48.05 46.19 46.8 47.61 39.057
2.23 1.65 1.79 2.05 1.97 1.62 1.94 1.72 1.43
77.79 76.01 74.26 71.79 76.05 76.93 70.11 79.64 73.38
0.604 0.632 0.591 0.641 0.632 0.600 0.668 0.598 0.532
0.0286 0.0217 0.0241 0.0285 0.0259 0.0210 0.0276 0.0215 0.0194
GO, unconverted gas oil; LG, light gases; GA, gasoline; sel, selectivity.
c5-c12
oil ratio) or regeneration conditions affected final results. Thus, carefully controlled experimental steps and procedures were considered for all experiments. The catalyst containing the small zeolites, so-called USSY-small, was found to be moderately more active than the USSY-large catalyst. It has to be stressed that pellets of both catalysts, USSY-small and USSY-large, also had very similar zeolite contents (close to 23 w t %). Therefore, the higher activity of USSY-small was con-
3058 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994
sistent with the diminished intraparticle diffusional controls in USSY-small. Gasoline fraction was defined as the total gas oil conversion minus the light gases and the coke. Light gases were considered to be in the C I - C ~ fraction. Results obtained show that the USSY-small catalyst had a higher selectivity for gasoline than the USSYlarge one. The percentage of the gasoline fraction obtained from the chromatograph for the USSY-small zeolites ranged around the low 50%, while the range for the USSY-large zeolites averaged around the middle 40%. Results obtained can be justified on the basis that under reduced diffusional constrains gas oil cracking should augment while gasoline cracking should remain relatively stable; gas oil molecules are significantly bigger than gasoline molecules. Also under these conditions there is an increased accessibility of gas oil to more isolated and stronger acid sites. The outcome is gasoline selectivities enhanced with the use of USSY-small catalyst. These results were confirmed by analyzing the selectivity to gasoline which is defined as the ratio between the gasoline fraction and the total gas oil conversion. A selectivity of around 0.65 for the USSY-small zeolite and a selectivity of 0.6 for the USSY-large zeolite were observed. Selectivity to gasoline shows, however, a tendency t o decrease with increasing temperatures. This is in accord with results obtained by other researchers suggesting that at higher temperatures hydrogen transfer reactions are depressed and consequently gasoline selectivity is reduced. Another important matter concerns changes of gasoline selectivity with residence time. In most cases it was observed that selectivity to gasoline had an optimum at 10 s with only a few conditions reaching this optimum at 5 s. These results suggest two very important facts: gasoline overcracking is essentially negligible; cracking reactions take place at the strongest acid sites, the first ones to be coked at small contact times, leading to higher fractions of light gases. The coke formed during each cracking run was analyzed using a coulometric analysis and thus on the basis of COz formed. The coke values obtained exhibited a consistent increase in coke formed as the temperature and the residence time in the reactor were increased. This observation held true for both USSY-small and USSY-large. It must be noted, however, that the coke produced by the USSY-large catalyst was invariably larger than the coke produced by the USSY-small catalyst. This result is very interesting given it demonstrates the higher selectivity of the USSY-small zeolites favoring the production of useful products such as gasoline rather than the production of excess coke. Moreover, lower coke yields, as shown consistently by USSY-small, may be a most valuable attribute while processing heavy feedstocks with strong coking tendencies. In this respect, these catalysts should provide a major advantage by simplifying the regeneration process without the need of catalyst coolers for catalyst regeneration (Mauleon and Courcelle, 1985). In order to correlate experimental findings with mechanistic considerations, such as H-transfer, the isoCJC4 olefin ratio was considered. This ratio, normally reported as an indicator of H-transfer (Sedran et al., 19921, was analyzed for various C/O, temperatures, and residence times (Figures 2 and 3). It was found that
s"
--
8
3
A
8 A 0 A
$ 0.4 G
W A 0 -t
0
li 0.3 -
d
u I
0
2
0
A
\
0.2 -
0 ussy SMALL TEMPERATURE
(OC)
0.1 500 525 550
0.0 0
I
I
1
2 3
I
3.6 3.6
I
I
4
5 6
I
1
1
1
1
7 8 9
)
10
t (SI
Figure 2. Iso-CdC1 olefins as a function of reaction time, temperature, and catalystloil (C/O) for USSY-small in the Riser Simulator.
-s
0.5
1 A
TEMPERATURE P C )
0
O.l 0.0
0
1
ll
2 3
500
52 5 550
4
5 t
c/o
i'jl 4.7
,
I
~
3.6
6
7 8 9
10
(5)
Figure 3. Iso-CdC4 olefins as a function of reaction time, temperature, and catalystloil (C/O) for USSY-large in the Riser Simulator.
both USSY-small and USSY-large catalysts, subject to extensive dealumination, displayed an increasing isoCJC4= ratio with residence time: from 0.25 t o 0.4. This indicates that low H-transfer, consistent with a unit cell size close to 24.28 A (Jacquinot et al., 19901, dominates at the onset of the reaction. Once sites with higher acid strength are coked, weaker sites influence in a much larger extent product distribution. Moreover, while comparing C d olefin yields for USSYsmall and USSY-large (Figure 4), an appreciable C4 olefin yield for USSY-small can be noticed. This consistent difference, at various temperatures, appears to suggest reduced H-transfer in the USSY-small compared with the USSY-large. Moreover, lower coke yields with the USSY-small are also in line with decreased H-transfer influence. As stated above, nonframework
Ind. Eng. Chem. Res., Vol. 33, No. 12,1994 3069
-::
SMALL
LARGE
Y
3.6
0
u'Y
500 525 TEMPERATURE
5 500 525 TEMPERATURE
550 (OCI
Figure 4. C4 olefins/gas oil converted as a function of reaction temperature for both USSY-large and USSY-small at various conditions studied in the Riser Simulator.
alumina could alter H-transfer. For instance, it is argued that nonframework alumina could coordinate with isolated sites in the network (Peters, 1992). Given that in smaller crystal zeolites gas oil concentration varies less markedly across the crystal diameter, the outer crystal region has a less significant role than in larger crystals and changes in product distribution can be expected. A major objective of this study was to determine whether the zeolite crystal size plays an important role in the product distribution of the gasoline fraction. In particular, it was important to determine if crystal size affected the total amount of aromatics, especially benzene, produced under the same reaction conditions. To investigate this matter data were obtained for the two catalysts, USSY-large and USSY-small, at different reaction conditions such as temperature, reaction time, and catalyst-to-oil ratio. Results demonstrate that the amount of benzene produced by the USSY-small zeolite was consistently smaller than the amount produced by the USSY-large zeolite. For example, as reported in Figure 5 at 500 "C the benzene was reduced from 2.4% to 1.45%the USSYsmall zeolites were used instead of the USSY-large ones. At 550 "C a similar trend was observed with reductions between 2.9% and 2%. This is a valuable result that favors application of USSY-small since benzene is one of the gasoline components considered as most harmful due t o its high toxicity. The total BTX aromatics, represented by benzene, toluene, p-xylene, and o-xylene was also determined. The results showed that, in most of the runs, the BTX aromatics produced by the USSY-small catalyst was lower than that by the USSY-large catalyst. For instance, as reported in Figure 6, the total amount of aromatics were reduced from 11.6% to 10% when the zeolite crystal sizes from USSY-large t o USSY-small were changed. Figures 5 and 6 also illustrate the changes of both benzene and BTX aromatics as a function of reaction temperature. In the case of USSY-small, at 550 "C the
550 (OC)
Figure 6. Benzene in gasoline as a function of reaction temperature for both USSY-large and USSY-small. No distinction is made of benzene levels at different C/O and residence times studied in the Riser Simulator.
m
/ u
I 10
500 . 525 TEMPERATURE
550
(OC)
Figure 6. BTX aromatics (summation of benzene, toluene, and xylene fractions) as a function of reaction temperature for both USSY-large and USSY-small. No distinction is made of benzene levels at different C/O and residence times studied in the Riser Simulator.
amount of benzene produced was 1.6% instead of 2%. For the same catalyst the amount of aromatics changed from 10% to 12.9% when the temperature was augmented from 500 "C to 550 "C. These results are also consistent with reduced H-transfer at higher temperatures. All these results are encouraging since higher yields of gasolines with lower content of aromatics and higher olefins are going to play a major role in the production of reformulated gasoline. In this sense the USSY-small zeolite could contribute significantly. Given the reduced diffusional constrains, USSY-small should have the following effects on FCC: (a) gas oil cracking rate (monomolecular reaction) augments given the increased accessibility of gas oil to stronger acid sites, and as a
3060 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 Table 12. Kinetic Constant Evaluation in Riser Simulator with Large Catalyst Correlation Matrix kin E1 k31' E31 k32' E32 hi" 1.0000 E1 -0.2120 1.0000 k31' 0.9994 -0.2120 1.0000 E31 -0.2084 -0.9984 -0.2078 1.0000 k32' 0.9179 -0.1039 0.9208 -0.1012 1.0000 E32 -0.0643 0.9271 0.0644 0.9345 0.0068 1.0000 kz" 0.7458 -0.1848 0.7352 -0.1839 0.5263 -0.0921 E2 -0.3509 0.6673 -0.3523 0.6483 -0.2602 0.3566 a 0.8195 -0.3113 0.8207 -0.3072 0.5949 -0.1534
kz"
E2
a
1.0000 -0.1676 0.5687
1.0000 -0.4199
1.0000
1.0000 -0.3763
1.0000
Variance of Residuals = 0.01937,45Degrees of Freedom Individual Confidence Limits for Each Parameter (Linear Hypothesis) kl" = 0.4272 x 1013 i 0 . 6 5 ~ lo6;E1 = 21 009.9i 3000 k31' = 0.1012 x loi4f 0.35 x lo6;E31 = 23 337.9 f 3500 k3z0 = 0.5504 x 10" f 0.2 x lo6; E32 = 20 934 f 11 000 kz" = 0.1337 x lo6 i 0.3x lo6;E2 = 17 461.4f 3000 a = 391 f 40
Table 13. Kinetic Constant Evaluation in Riser Simulator with Small Catalyst Correlation Matrix 1.0000 -0.3545 0.9987 -0.3504 0.9104 0.1363 0.2333 -0.3663 0.6367
1.0000 -0.3540 0.9960 -0.1841 0.7609 -0.0387 0.6208 -0.3369
1.0000 -0.3480 0.9104 0.1375 0.2175 -0.3679 0.6423
1.0000 -0.1799 0.7699 -0.0410 0.5983 -0.3329
1.0000 0.2955 0.1531 -0.2719 0.3634
1.0000 -0.0243 0.0619 -0.0010
1.0000 -0.0598 -0.2826
Variance of Residuals = 0.02766,45Degrees of Freedom Individual Confidence Limits for Each Parameter (Linear Hypothesis) k1" = 0.9419 x 1013f 0.45 x lo7;E l2 = 22 795.3 f 2000 k3i0 = 0.3769 x loi5 f 0.20 x lo7;E31 = 29 900.5f 2000 k32' = 0.1070 x 1015f 0.12 x lo5;E32 = 34 824.8f 13 000 k2" = 0.1093 x lo4 i 0.035: z- = 11 402.3 f 3500 ,E &= 526.56i 44
result gasoline selectivities are enhanced; (b) rate of bimolecular reactions such as H-transfer reaction are strongly depressed, due to the low density of acid sites. Thus, the overall effect is gasoline containing more olefins and less aromatics. Consistent with this much less coke is formed. 5. Kinetic Constants As already described in the modeling section the fourlump model, highlighting unconverted gas oil, light gases, gasoline, and coke, was used t o determine the kinetic constants of two catalysts: USSY-large and USSY-small. The four model equations under the particular conditions of the Riser Simulator were described with eqs 9, 12, 13, and 14. An Arrenhius type dependence with respect to temperature was considered for the various kinetic parameters. An exponential decay function, with activity decaying as a function of coke on catalyst, was also employed. While kz = kzl k22 reflects gasoline overcracking, these two constants have in practice very different magnitudes,kzl >> k22, and consequently, k22 was the only overcracking constant considered further in the analysis. In order to provide better parameter estimates, the kinetic parameters were subject to reparametrization as proposed by Blasetti (1994). Parameters were adjusted using a weighted least-squares algorithm for nonlinear parameter estimation (Marquardt, 19631, regressing simultaneously on the unconverted gas oil,
+
the gasoline, the coke, and the light gases lumps. In this way nine kinetic parameters were simultaneously adjusted. Tables 12 and 13 report the evaluated kinetic constants with the cross-correlationmatrix and the respective parameter spans. It can be observed that in the cross-correlation matrix most of the coefficients remain in the low level with only a few exceptions. For example, for the case of the USSY-small catalyst (Table 13)the correlation matrix coefficients point toward some degree of correlation between k1" and k3Io, as well as between E1 and E31. It appears that in these cases the expected correlation was not completely eliminated. Moreover, for the case of USSY-large catalyst, the crosscorrelation matrix gives similar good results providing low correlation between parameters with only a few exceptions. Again an unbroken correlation between k1" and 1231" parameters on one side and E1 and E31 parameters on the other was noticed. It is also interesting to mention that a review of Tables 12 and 13 provides evidence that for both USSYsmall and USSY-large catalysts all constants of the kinetic model are obtained with quite restricted spans for the 95% confidence interval. It has to be pointed out that in particular the activation energies determined for the reaction steps of gas oil conversion into gasoline and gas oil conversion into light gases showed very restricted spans. The adequacy of the "coke-on-catalyst" model was confirmed by comparing the predicted gas oil conver-
Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 3061
a 3-
-2 -3
0
I
I
0
i
0
0
> 0
10
20
30
40
50
60
Experimental Observations 3 ,
1
?2 v)
-Q
v)
a 0
I
0:
-2
L
-l -3
0
10
20
30
40
50
60
Experimental Observations Figure 7. Distribution of {residualshtandard deviation) for various experimental observations (unconverted gas oil, gasoline, light gases) using the four-lump model in experiments with (a) USSY-large and (b) USSY-small.
sions, gasoline and light gases lumps with the observed gas oil conversions, and gasoline and light gases lumps. Typical random deviations of residuals, reported in Figure 7, were found for the four lumps under consideration. These deviations with most of the observations in the f2(residuals/standard deviation} range confirm no outlier observations and model appropriateness (Mendenhall and Sincich, 1992). Concerning k l / ( k l f k 3 1 f k32) and k32/(kl f k31 k 3 2 ) ratios, they show the higher selectivity of the USSYsmall zeolites for gasoline (69.4% vs 65%) and lower selectivities for coke (0.38% vs 0.86%). Moreover, the low value of the kz constant points toward little contribution of overcracking for all conditions studied for both USSY-small and USSY-large zeolites. On the strength of these results and various considerations described above it is valuable to compare the energies of activation obtained for the USSY-small and the USSY-large and data for commercial USY catalysts reported in the technical literature. The activation energies for the two catalysts, USSY-small and USSYlarge, revealed somewhat higher activation energies for the USSY-small zeolite with respect to the USSY-large (Tables 12 and 13). Moreover, the activation energies for the USSY-large are in the range of the activation energies observed with Octacat (18 000-22 000 caVmol), a commercial USY catalyst (de Lasa and Kraemer,
+
1992). These are important results given they are in accordance with the anticipated idcrease of activation energies with a reduction of zeolite crystal size. In fact, it can be expected that smaller crystal sizes will promote smaller diffusional controls and, consequently, higher activation energies. It must be pointed out that the activation energies for the coke-on-catalystdecay model are appropriate, given the statistical analysis developed. Also these parameters are obtained in the context of the coke-on-catalystmodel which provides, when compared with other more empirical approaches (e.g., time on stream), a more sound description of catalyst deactivation phenomena.
6. Conclusions The main objective of this study was t o determine whether the crystal size of the Y zeolites, the active phase of FCC catalysts, can affect gasoline product distribution. Zeolites were synthesized (USSY-large and USSYsmall) with mean crystal sizes of 0.41 and 0.13 pm respectively and SUAl ratio of 2.4. The above catalysts were activated and stabilized using ion exchange and steam calcination to obtain ultrastable Y zeolites. The zeolites were pelletized using spray-drying techniques. Characterization of the catalysts was carried out using XRD, SEM-EDX, BET, TPD, and electron scanning microscopy t o follow the physical and chemical changes in the zeolite structure during each treatment period. The resulting FCC catalyst had a mean particle size of approximately 45-60 pm and had a spherical shape ideal for a fluidized bed reactor. Reaction results obtained in the Riser Simulator indicate a high catalytic activity for both USSY-small and USSY-large catalysts. Higher selectivity to gasoline was also favored in the USSY-small zeolites by approximately 5 wt %, and this is consistent with reduced intraparticle diffusional controls for gas oil catalytic cracking (monomolecular reaction). Results obtained also show that the total amount of aromatics produced by the USSY-small zeolites was smaller than the one obtained for the USSY-large zeolites. In particular, the amount of benzene was found to be consistently smaller in the USSY-small zeolites. Furthermore, a larger amount of Cq olefins was generally produced by the USSY-small zeolites and this effect was assigned to very limited H-transfer in the USSY-small. Valuable kinetic modeling and good lump predictions were obtained with a four-lump model highlighting unconverted gas oil, gasoline, light gases, and coke and relating catalyst deactivation on coke-on-catalyst concentration. This model also provides important insights on the energies of activations for each reaction step with higher energies of activation for the USSY-small zeolites with respect to the USSY-large zeolites suggesting lower intracrystal diffusional controls.
Acknowledgment We would like to acknowledge the financial support received during the development of this study by the Natural Sciences and Engineering Research Council of Canada (NSERC) through the Strategic Grant Program.
Nomenclature A = symbol adopted for gas oil in the four-lump model B = symbol adopted for gasoline in the four-lump model
3062 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 b = width of the q u a r t z peak B' = width of the peak of the zeolite XRD analyzed C = symbol adopted for light gases in the four-lump model CA = concentration of gas oil (g-moYcm3) CB = concentration of gasoline (g-moYcm3) D = symbol adopted for coke in the four-lump model E = energy of activation (caYg-mol) K = constant in eq 1 kl = individual kinetic constant for gas oil cracking t o gasoline in the four-lump model (m6/(kmol.kgcat*s)) k21 = individual kinetic constant for gasoline cracking t o light gases in the four-lump model (m3/(kgc,t*s)) k z z = individual kinetic constant for gas oil cracking to coke i n the four-lump model (m6/(kmol.kgcat*s)) ks1= individual kinetic constant for gas oil cracking to light gases in the four-lump model (m6/(kmol-kg,,t*s)) k32 = individual kinetic constant for gasoline cracking t o coke in the four-lump model (m3/(kg,,t.s)) k2 = summation of kzl and k22 (m3/(kgCat.s)) k3 = summation of k31 and k32 (m6/(kmol.kgcat.s)) M A = molecular weight of gas oil (g/g-mol) M B = molecular weight of gasoline (g/g-mol) MW = average molecular weight of the gas mixture (g/gmol) m, = mass of catalyst (g) R = universal gas constant (atmcm3/(g-mol-K)) r-A = rate of gas oil cracking (g-moY(g of cabs)) rB = rate of gasoline formation (g-moV(g of cabs)) t = time on stream (s) u1 = stoichiometric coefficient for gasoline formation from gas oil (MA/MB) VT = internal volume of the reactor (mL) W T = total weight of hydrocarbons in the reactor (g) YA= weight fraction of the gas oil YB= weight fraction of the gasoline Yc = weight fraction of light gases Y D= weight fraction of coke Superscripts O
= refers to the preexponential factor of the kinetic constant
Greek S y m b o l s
a = exponential decay function /3 = diffraction broadening 6 = Bragg angle
1 = wavelength (A) d = fraction of the active sites
Literature Cited Blasetti, A. Multitubular Reactor for Catalytic Cracking of Hydrocarbons. Design and Kinetic Modeling. Ph.D. Dissertation, University of Western Ontario, London, ON, Canada, 1994. Coxson, P. G.; Bischoff, K. B. Lumping strategy. 1. Introductory techniques and applications of cluster analysis. Znd. Eng. Chem. Res. 1987,26,1239-1248. de Lasa, H.I. Novel Riser Reactor Simulator. Canadian Patent 1,284,017, 1991; USA Patent 5,102,628, 1992. de Lasa, H.;Kraemer, D. Novel Techniques for FCC Catalyst Selection and Testing. In Chemical Reactor Technology for Environmentally Safe Reactors and Products; de Lasa, H., Dogu, G., Ravella A., Eds.; NATO Series 225; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1992; pp 71-131. Farag, H. Catalytic Cracking of Hydrocarbons with Novel Metal Traps. Ph.D. Dissertation, University of Western Ontario, London, ON, Canada, 1993.
Farag, H.; Ng, S.; de Lasa, H. Kinetic Modeling of Catalytic Cracking Using in Situ Metal Traps (FCCT) to Prevent Metal Contaminants Effects. Znd. Eng. Chem. Res. 1993,32,10711080. Gianetto, A. Novel Cracking Catalyst for Reformulated Gasoline. Master of Engineering Science Thesis, University of Western Ontario, 1993. Jacob, S. H.; Gross, B.; Voltz, S. E.; Weekman, V. M. A lumping and reaction scheme for catalytic cracking. M C H E J. 1976,22, 701-713. Jacquinot, E.; Mendes, A.; Raatz, F.; Marcilly, C.; Ribeiro, F.; Caeiro, J. Catalytic Properties in Cyclohexene of Modified HY Zeolites. Appl. Catal. 1990,60,101-117. Kraemer, D. W. Catalytic cracking in a novel riser simulator: design and testing. M.E.Sc. Thesis, The University of Western Ontario, London, ON, Canada, 1987. Kraemer, D. W. Modeling catalytic cracking in a novel riser simulator. Ph.D. Dissertation, University of Western Ontario, London, ON, Canada, 1991. Kraemer, D. W.; de Lasa, H. I. Catalytic cracking of hydrocarbons in a Riser Simulator. Znd. Eng. Chem. Res. 1988,27(ll),20022008. Leviosiforich, M. Handbook of X-ray Analysis of Polycrystalline Materials; Consultants Bureau: New York, 1964. Marquardt, D. W. A n algorithm for least squares estimation of nonlinear parameters. J. SOC.Znd. Appl. Math. 1963,11, 431. Mauleon, J. L.;Courcelle, J. C. FCC heat balance critical for heavy fuels. Oil Gas J. 1985,Oct 21, 64-70. Mendenhall, W.; Sincich, T. Statistics for Engineers and the Sciences; MacMillan Publishing Corp.: New York, 1992. Pekediz, A. Oxidative Coupling of Methane in a Novel Riser Simulator with Staged Oxygen Injection. Ph.D. Dissertation, University of Western Ontario, London, ON, Canada, 1994. Pekediz, A.; Kraemer, D.; Chabot, J.; de Lasa, H. I. Mixing patterns in a novel riser simulator. In Chemical Reactor Technology for Environmentally Safe Reactors and Products; de Lasa, H., Dogu, G., Ravella, A., Eds.; NATO Series 225; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1992; pp 133-146. Peters, A. W.; Cheung, W. C.; Roberie, T. G. Scientific Aspects of Novel Catalysts in FCC. In Chemical Reactor Technology for Environmentally Safe Reactors and Products; de Lasa, H., Dogu, G., Ravella, A., Eds.; NATO Series 225; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1992; pp 51-69. Rajagopalan, K.; Peters, A. W.; Edwards, G. C. Influence of zeolite particle size on selectivity during fluid catalytic cracking. Appl. Catal. 1986,23,69-80. Seddon, D.Reformulated gasoline, opportunities for new catalyst technology. Catal. Today 1992,15, 1-21. Sedran, U.; Kraemer, D.; de Lasa, H. Evaluacion de Catalizadores y Alimentaciones de Cracking Catalitico. Proceedings of the ZberoAmerican Conference on Catalysis, Segovia, Spain, 1992. Stokes, G. M.; Wear, C. C.; Suarez, W.; Young, G. W. Reformulated gasoline will change FCC operations and catalysts. Oil Gas J. 1990,July 2. Weekman, V. W. A model of catalytic cracking conversion in fxed, moving and fluid-bed reactors. Znd. Eng. Chem. Process Des. Dev. 1968,7,90-95. Wei, J.; Kuo, J. C. W. A lumping analysis in monomolecular reaction systems. Analysis of the exactly lumpable system. Znd. Eng. Chem. Fundam. 1969,8(l), 114-123. Received for review March 25, 1994 Revised manuscript received September 19, 1994 Accepted September 30, 1994@ Abstract published in Advance A C S Abstracts, November 1, 1994. @