You are on page 1of 15

39

JEN-RAY QiANG, ~aN~~~~~N~ SHEU, Y~N~"M~N~ CMEMG and J~N~~C#~~~ b&J*


Refining & Manufacturing Research Center, Chinese Petroleum Carporation
Chia-Yi, Taiwan 6DO36, Republic of China,

(Received 9 October 1986, accepted 20 March 1987)

A semj-emp~r~caJ decay model of taluene d~spr~por~~~nat~~n was derived and


verified by kinetic studies, The activities of c~~erc~aJ solid acid cataJysts
for toluene d~~pr~p~rti~n~~i~n were tested in an up-f‘iowfixed-bed reactor under
atmospheric pressure and at 743-803 K with WNSV = 0.76, Coke was f@.indto form
from different kinds of precursor, such as p#lycyclic-aroN~atichydrocarbons and
poly-olefins, Computer simulation was used to obtain the optimal operative
temperature for a moving-bed reactor, It was found that the higher the catalyst
flow rate, the higher the temperature required for optimal conversion under a
constant feed flow rate, and also the hSgher the s~e~d~-stat~ c~~versi~~ of
toiuene. S~mffJatj~~of the proposed decay made7 gave Jess than 5X deviation from
actual commercial plant operation.

A surging demand for benzene and xylene as raw materials for synthetic resins
and fibers Sn recent years has st~muJat~d severaf de~eJ~pme~ts in converting
taluene to benzene and xylene via disproportionat~on or dea~k~lat~on processes,
These d~spr~portionation or dealkyla~ion processes are normally carried nut in
the vapor phase aver solid acid catalysts at elevated temperatures. The dis-
proportionat~o~ of toluen~ produces benzene and xyfene as the main products~
Some by-p~oduc~s~ such as light paraffins, C9* alkylben~enes and pofycyclic-
aromatic compounds, are also formed. Mordenite C1,21, rare-earth exchanged X
teolite [3,4J and cation-exchanged Y zeolite C5,61 all possess high activity for
the reaction, unfortunately, many of these catalysts show low selectivity for
disproportionation as a result of crack-fngand hyd~odealkylation, F~r~hermore~
their activities decline quickly due to coke deposition and all but cease within
a fe% hours, ~~co~~ngl~~ deve?o~en~ of new toluene d~sproportionatjon cata-
lysts has shifted to the use of composite catalysts [7-101. Oliver and Inove
summar~2ed industrial toluene disproportio~ation research and pointed out that
vapor phase systems with solid catalysts have significant economic advantages
tll3. Successful c~erc~a~ processes have been claimed by S~n~~ajr/~~la~tic
Richfield [113 and Toray rndustrial/~niversaJOil Products [12,133. SeJective
40

toluene disproportionation (STDP) and liquid phase low temperature dis-


proportionation (LTD) processes have also been developed by the Mobil Company
with modified ZSM-5 catalysts [14-171.
From a literature survey, coke formation on catalysts can take several forms
and be caused by different mechanisms [18]. Eberly et al. suggested that coke
formation involves initial adsorption of hydrocarbon followed by chemical
reactions on the catalyst surface [19]. The chemical reactions here may include
condensation followed by hydrogen elimination. Inits simplest form this may
occur by reactions in parallel or in series to the main reaction. Early work by
Blue and Engle suggested that olefins are largely responsible for the coke
formation at low conversion and that aromatic compounds contribute significantly
to the formation at higher conversion. In any event, the precursors to coke
formation are still not wholly understood [20].
Studies of the deactivation of catalysts provide useful information for
designing new catalysts and/or reactors, and help to optimize reaction conditions

in industrial processes. A number of methods have been developed to deal with


catalyst deactivation and these have met with varying degrees of success [19,
21-313. Although they are formally different, the mathematical structure of
these methods is generally similar.
In this work, kinetic studies were carried out in a fixed-bed reactor for
parameter estimation and to establish the deactivation model. Optimization of
operating conditions for a commercial moving-bed process was also worked out
by computer simulation.

EXPERIMENTAL
A once-through up-flow fixed-bed reactor was used for the activity tests. The
reactor is a vertical stainless-steel tube with an inside diameter of 1.8 cm.
The catalyst bed, packed with 30 g of catalyst (3.5 mm dia. spheres), was first
heated to the reaction temperature under a nitrogen flow of 50 ml min -' to
remove any contamination. The reactions were then studied under the following
conditions: temperature, 743-803 K; pressure, atmospheric; WHSV, 0.76. During
each run, samples were taken at intervals and analyzed by an FID gas chromato-
graph with a 8 ft x l/8 in. o.d. stainless-steel column packed with 5% DIDP/5%
Bentone 34 on 100/120 Chromosorb W HP under the following conditions: oven
-1
temperature, 373 K; flow rate, 20 ml min N2; injection volume, 0.5 ul. Exhaust
gas from the reaction and the coke on the used catalyst were subjected to PONA
(paraffin, olefin, naphthene and aromatics) and elemental analyses respectively
for material balance calculation. The minor components in the effluent were
analyzed and identified by GC-Mass spectra and PONA tests. Three kinds of
commercial catalysts, Cat-A, Cat-B, and Cat-C were used in the performance tests.
The characteristics of these catalysts are shown in Table 1. The feedstock of
toluene used in the experiments is a product of the Chinese Petroleum Corporation
and has a purity of 99.9% +.
Cat-A Cat-B &at-&
Catalyst

Composition

Si02/wtX 45 49.5 49.5

;;N$ 520.5 48.8


0.2 49.0
0.2
2
Cr203/wt% 0.1 0.1

Fe203/wt% 0.2
T~~~/~~~ 1.0
ia203/wt?% l-2 0.5 0.3

Physical properties

diameter/mm 3.8 (sphere) 3.5 (sphere) 3.5 (sphered


bulk density/g cm -3 0.92 0.93 0.90
surface area/m* g -1 140 104 104
pore volume/cm3 g-l 0.25 0.35 0.35

aAbstracted from technical data of the Engelhard and Kali-Chemie AG Carp.

TIME ON STREAM, hi-

FIGURE 1 Test for external diffusion resistance; catalyst: Cat-B; reaction


conditions: 773 K, and atmospheric pressure.
0 0.5 1.0 1.5 2.0 2.5 3.0

TIME ON STREAM, hr

FIGURE 2 Test for internal diffusion resistance; catalyst: Cat-B; reaction


conditions: 773 K, atmospheric pressure and WHSV = 0.76.

RESULTS AND DISCUSSIONS


Mass transfer limitations
In order to test the external mass transfer limitation of the reaction

system, the reactions were carried out at constant WHSV with different feed

rates and different amounts of catalysts. Figure 1 shows the plot of toluene
conversion vs. time on stream for the toluene disproportionation reaction over

the commercial catalyst Cat-B. Since the curves obtained for two different feed
rates are coincident, it can be concluded that the external mass transfer
limitation is negligible during the test. The results also imply that the
deactivation of the catalyst does not affect the occurrence of external mass
transfer limitation.
Another method of evaluation of the external mass transfer limitation is to
calculate the partial pressure of gas at the surface of the catalyst particles

in fixed bed reactors. This method was developed by Yoshida et al. C32J. Using
this method, in the present reaction system, the modified Reynolds number was
1.42, the rate number varied from 3.23 x 10m3 to 8.0 x 1o-4 in the reaction
period, and the Schmidt number was estimated to be 0.68. By using these data
the partial pressure gradients were estimated and found to vary from 3.6 x low3
to 0.9 x 10-3, indicating that there was no significant external mass transfer
limitation in the given reaction system.
In order to test intraparticle mass transfer limitation, the reactions were
also carried out over the same kind of catalysts with various diameters. The
results are shown in Figure 2, where the curves arecoincidentinitially but
43

diverge slightly'towards the end of the run. This indicates that there is no
significant intraparticle mass transfer limitation for the evaluated catalysts
in the following performance tests. The slight divergence of the plots indicates
a little intraparticle mass transfer limitation due to slight pore-plugging
which occurs only after severe coking. Since the operational conditions for
c~mercial application and our experimental mode? did not cause severe coking,
the intraparticle mass transfer limitation can be neglected in the present work.

o Cat -A
r Cat--B
l Cat-C

I I I I
0.5 I.0 1.5 2.0

TIME ON STREAM, hr

FIGURE 3 Performance tests of commercial catalysts for toluene dispr~~~rti~na-


tion; reaction conditions; 773 K, atmospheric pressure and WHS\I= 0.76.

@ic model of the tolupne disproportionation reaction


In the reaction of toluene disproportionation over a solid acid catalyst,
catalyst deactivation may be caused by both the reactants and the products. A
simple mathema~ica7 model based on a single irreversible reaction and n-th order
concentration-independentdeactivation can be used to describe the fixed-bed
reaction system C301.

Y-,. = kCTma
(1)
- $ = kda"
(2)
44

intercept =ln kr

0 0.5 1.0 1.5 2.0

TIME ON STREAM, hr

FIGURE 4 Plots of In In CT G/CT vs. time on stream for toluene disproportiona-


tion over Cat-A, Cat-B and tat-C at 773 K and atmospheric pressure.

where yT is reaction rate; CT is the concentration of reactant; m and n are the


orders of reaction and catalyst deactivation respectively; k and kd are the
reaction rate constant and deactivation rate constant respectively; "a" is a
dimensionless fractional catalyst activity number and is defined as

a = rate at which pellets convert reactant T


rate at which fresh pellets convert reactant T

Figure 3 shows the performance tests for toluene disproportionation over

three different kinds of commercial catalysts. A material balance of the system


gave 98-99% of effluent to feedstock, indicating that this packed system was
adequate for the activity test. The selectivity for the formation of benzene
plus xylene was found to be in the range of 96-99% during the run.
Using a least-squares method, the best agreement between theoretical and
experimental data was obtained with a first-order reaction and first-order
catalyst deactivation. Equations (I) and (2) then become

yT = kCTa
(3)

_ da
= kda
X (4)
45

Integrating equation (4), yields

(5)
a = a eekdt
0

and at the start of the reaction, a0 = 1, therefore (5) becomes

For the constant plug flow of fluid, the performance expression combined with

the rate of equation (3) gives

w'FT,O =
J
'T
o
dXT=
YT s
'T

0
dXT
kaCT
-1
= kaCT o
,
s
'T

'T,O
dCT

CT
(7)

where, W: weight of catalyst, gram; FT o: feed rate of reactant, (mol 1-l).


(gram of feed h-l); XT: fractional con$ersion of reactant T; CT o: initial con-
-1 ,
centration of reactant, mol 1 .
Integrating equation (7) and replacing "a" by the expression of equation (6),
gives

(8)

Rearranging equation (8) gives

In In $$ = In (k-r) - kdt (9)

where T = t/WHSV is weight hourly space time.

Figure 4 shows the results of the kinetic experiments to evaluate equations


(3) and (4). The reaction and catalyst deactivation rate constants were ob-
tained from the intercepts and slopes of each straight line. The results are
shown in Table 2, where catalytic activity is in the order Cat-A >Cat-B>Cat-C
and catalytic deactivation is in the order Cat-A z Cat-B > Cat-C.

Effect of reaction temperature on catalyst deactivation


The activity tests of commercial Cat-B were carried out in a fixed-bed
reactor under atmospheric pressure with IWHSV = 0.76. The reaction temperatures

were varied from 743 to 803 K. Figure 5 shows the conversion of toluene VS.

time on stream. It indicates that the conversion decreases monotonically with

time on stream.

The reaction rate constant k and catalyst deactivation rate constant kd are
expressed in the Arrhenius form,
46

TABLE 2
Reaction and catalyst deactivation rate constants of commercial CatalyStS at

773 K.

Catalyst Cat-A Cat-B Cat-C

k /h-l 0.466 0.338 0.256


kd fh" 0.450 0.462 0.408

TIME ON STREAM, hr

FIGURE 5 Effect of reaction temperature on catalyst deactivation over Cat-B;


reaction conditions: atmospheric pressure and WHSV = 0.76.

k = k’ e-Ea/RT (10)
0

-Ed/R7
kd = kfl,o e (12)

where E, and Ed are the activation energies of the main reaction and catalyst
deactivation respectively.
Plots of In k and In kd against l/T are shown in Figure 6. As can be seen,
astraightlinewasobtainedforthcplotoflnkvs.l/Tandacurveforlnkdagainst
l/T. Therefore, In k against t/T follows an Arrhenius law, whereas In kd against
t/T does not. The curve of ln kd against l/T can be approximated to two straight
line plots, as shown in Figure 6, with one straight line plot representing a
higher activation energy than the other.
1.25 1.30 1.35

1 /TX mOCK-'1

FIGURE 6 Arrhenius plot of the toluene dispraportionation reaction and catalyst

(0.061) (0.049) (0.047) (0.057) c 0.007)

FIGLIRE 7 Precursors of coke formation in the eff?uent as found by GC-!&ass anal-


ysis; catalyst: Cat-B, temperature: 773 K, time on stream: 1.5 h, WHSV = 0.76,
pressure: atmospheric; numbers in parenthesis indicate relative weight per cent,
absolute yield of naphthalene is 0.3 wt%.
A possible explanation for the two different activation energies iS that the

coke formation in the two reaction regions is due to two different kinds of pre-
cursor. From GC-Mass analysis of the effluent, minor components were identified

and accounted for as shown in Figure 7. The majority of components are poly-

cyclic-aromatic hydrocarbons of less than four rings. Trace amounts of non-


aromatic compounds, such as methyl ethyl ketone, methyl butyl ketone, methyl-
cyclopentane, cyclohexane and methyl-cyclohexane were also detected in the
effluent. From this information it is presumed that the main cause of catalyst
deactivation is coking due to the formation of polycyclic-aromatic hydrocarbons

and poly-olefins. Poly-olefins are formed as by-products during polycyclic-

aromatic hydrocarbon formation. Appleby postulated that interactions involving

benzene and naphthalene might be taken as representative of important reaction


paths in coke formation and that the acidic property of the catalyst may be
responsible far coke formation, presumably via the formation of carbonium ions
generated from various aromatic compounds [18].
Based on the above discussion, a semi-empirical catalyst deactivation model
is derived in the following way:

A+S + (12)
e SA

o+s e + sO (13)

se = St - s* - so (14)

dSe
-[kACA + kGCC]Se
TE= (15)

where A and 0 represent the different kinds of coking precursor, polycyclic


aromatic hydrocarbons and poly-olefins respectively; CA, CO are the concent-
rations of A and 0; Se, St, SA and SO are the effective sites, total active
sites, A-adsorbed sites and O-adsorbed sites respectively; kA and kO are the
rate constants of equations (12) and (13); Se/St = a, and a=l for t=O. Substi-
tuting the Arrhenius law into equation (15), gives

da -E '/RT
-Ck;CAe A + k" C'e-Et?RT] a
X= 0 0 (16)

where E;\ and E;1 are the adsorption activation energy of the precursors A and 0
respectively.
Since the numbers of aromatic rings and methyl groups do not change during the
disproportionation reaction, we assume that the concentrations of precursors A
and 0 are independent of the disproportionation reaction, and therefore C
and
A
Co can be expressed in terms of the empirical equations
49

(17)
CA = kP, (CT,D)ml (,)~~,-EP~/RT

CO = kP2(CT,0)mZ (T) "Ze-EP2/Rr

where E,,,and Ep2 represent the activation energy of the formation of precursors
A and 0 respectively; 'Iis space time.
In our case, CT,O and T are constant, therefore substituting equations (17) and
(18) inti equation (16) gives

_EAjRT -EO/RT
da = -[KAe + KDe la ( 19)
n

where KA = kprki (CT O)ml (~)~l, K. = kp2k; (CT D)mz(~~2,EA=~~ f EP,,ED=E,j+ EP2
Therefore, in the catalyst deactivation mathemaiical model, the temperature-
dependent kd should be taken as

-EA/RT -ED/RT
kd = KAe + Koe (20)

Parameter estimation and computer simulation


The estimation of frequency factors, k;, KA and I$,,and activation energy
Ea, EA and E. in equations (10) and (20) could be achieved directly by a non-
linear regression technique [33]. However, the plotting of In In CT O/CT vs.
time on stream is a much more straightforward method of estimating &e reaction
rate constant k and catalyst deactivation rate constant kd. The frequency factor
and activation energy E, were obtained from an Arr~~nius plot and found to
k:
-I
be 1607 h-l and 54.3 kJ mol respectively. The other parameters, KA, Kg, EA and
EO were obtained by the Gauss-Newton method C34] and calculated to be 4.48 x
IO" h-' , 87.0 h-l , 780.6 kJ mol-' and 38.0 kJ mol-' respectively. These kinetic
data were used in a computer simulation of the toluene disproportionation
reaction over Cat-B, The variance between experimental data and computer-
simulated data of toluene conversion at various WHSV is shown in Figure 8. Due
to the fact that, for a given value of WHSV, the two curves (one representing
the computer-simulated data and the other representing the experjmental data)
are almost superimposed. The proposed mathematical model and estimated para-
meters are adequate for the optimization of the toluene disproportion~tion
reaction in the WHSV range 0.50-1.00.

Optimization of the toluene disproport~onation reaction


Based on the above results and discussions, the optimization of the operating
temperature for a moving-bed reactor in a c~mercia~ plant can be carried out
by computer simulation. Since the disproportionation of tofuene is a slightly
50

0 msv=o.50
50 A wHSV=O.76
OS n wI-Isv= 1.00
*
oz 40
m

z
2 30
2
s
w 20

wz
2 10 -experiment
a ----computer sw3lat ion

0 0.5 1.0 1.5 2.0

TIhN2 ON STREAM, hi-

FIGURE 8 Comparjso~ between ~xperjmenta~ observations


and computer simulation
at various WHSV values for Cat-B; reaction conditions:
773 K and atmospheric

38

34

30
26

22

18

14 - reactant flow rate=100 rr? /hr


I I I I I
763 783 803 823 843

REACTION TEMPERATURE, K

FIGURE 9 Optimization of tofuene disproportionation by computer simulation for


a commercial moving-bed reactor with A = 23.62 III'
and 1.= 5.481 m using Cat-B as
the catalyst,
51

and the temperature drop of the


endothermic reaction (AH = 0.79 kJ mol-1) C91,
reactor is slight, the average t~perature Of the reactor can be used to

represent the operating temperature.


For the moving-bed reactor, another mathematical model can be det+ed by the
introduction of Q, and Q,, which represent the volumetric flow rate of reactant
and catalystrespectively.
Considera steady-statemoving-bed reactor of length

L through which fluid is flowing at a constant velocity Q,. The Composition of

the fluid varies from Point to point along the flow path; Consequently, the

material balance for a tY?aCtant component must be accounted for in a differential


element of volume dV. Thus,

'LCT,i - QLCr,f = yTdV (21)

-Q,dC, = vTAdl (22f

Rearranging and replacing yT with equation(3) gives

(23)

where CT i, CT f and A are the input concentration and output concentration of


the reactant a;d the cross-sectional area of the differential volume dV respecti-

vely; a(T) is the fractional catalyst activity and a function of the reactor
length 1.

For the deactivation, equation (4) can be rederived as follows:

9s = $ = catalystmoving rate
A- (24)

dt = A dl
F (251

Substituting equation (251 into equation (4) gives

da
= ($ 1 kda(ll
di (26)

Incorporating equations (lOI and (20) into equations (23) and (26) with the
boundary conditions, l=O, a=1 and CT=CT Q, the optimal operation temperature for
certain Qs can be calculated by compute; simulation. Theresults are shown in
Figure 9. These indicate that the higher the catalyst flow rate, the higher the
Optimal operation temperature for toluene conversion at constant Q, and the
higher the steady-state conversion_
52

CONCLUSION 3
The optimal operation temperatures for the reaction at constant QL* loo m /
,.,,.
with values of ~~ at 110, 190, 270 and 350 m3/hr were found to be 7B8y 803,
813 and 825 K, and gave a toluene conversion Of 28.5%, 34.6%~ 37.5% and 39*5%

respectively. From the simulation data shown in Figure 9, we ak0 brunt that ill
the reaction conditions: temp. = 758 K, WHSV = 0.76, QL = 100 m3/hr and Q, =

190 m3/hr the reaction gave 28.3% toluene conversion, while the actual
commercial plant running at the same conditions gave 27.0% conversion, The
difference between the semi-empirical model and plant operation for toluene
conversion is less than 5%.

In COnCluSiOn, our proposed semi-empirical catalyst decay model has been


proved to be adequate for the simulation and optimization of toluene dispro-
portionation in a moving-bed reactor.

ACKNOWLEDGEMENTS

The authors wish to express their sincere thanks to Dr. J.L. Fan for her help
in GC-Mass analysis and Mr. D.S. Hwang of the xylene plant at the Kaoshiung
Refinery of the Chinese Petroleum Corporation for many fruitful discussions.

REFERENCES
H.A. Benesi, J. Catal., 8 (1967) 368.
T. Yashima, H. Moslehi and N. Hara, Bull. Japan Petrol. Inst., 12 (1970) 106.
H. Matsumoto and Y. Morita, Bull. Japan Petrol. Inst., 1C (1968) 8.
P.B. Venuto, L.A.Hamilton, P.S.Landis and J.J. Wise, J. Catal., 5 (1966) 81.
P.A. Jacobs, H.E.Leeman and J.8. Uytterhoeven, J. Catal., 33 (1974) 31.
K.M. Wang and J.H. Lansford, J. Catal., 24 (1972) 262.
J.C. Wu and L.J. Leu, Appl. Catal., 7 (1983) 283.
L.C. Gitber'let and R.J. Bertolacini, U.S. Pat. 3,548,020 (1970).
L.E. Aneke, L.A.Gerritsen and W.A.de Jong, J. Catal., 59 (1979) 26.
L.E. Aneke, L.A.Gerritsen, J. Eilers and R. Trion, J. Catal., 59 (1979) 37.
E.D.Oliver and T. Inove, Stanford Research Institute Handbook No. 30A (1970).
T. Ponder, Hydrocarbon Processing, Nov. (1979) 141.
S. Otani, S. Matsuoka and M. Sato, Japan Chem. Quart., 4 (1968) 16.
P. Grandio, F.H. Schneider, A.B. Schwartz and J.J. Wise, Oil Gas J., 69
(1971) 62.
W.W. Kaeding, C. Chu, 1.8. Young and S.A. Butter, J. Catal., 69 (1981) 392.
W.W. Kaeding and S.A.Butter, U.S. Pat. 3,911,041 (1975).
W.W. Kaeding and L.B. Young, U.S. Pat. 4,034,053 (1977).
W.G. Appleby, J.W. Gibson and G.M. Good, Ind. Eng. Chem. Process Des, Dev.,
1 (1962) 102.
P.E. Eber?y, C.N. Kimberlin, W.H. Miller and H.V. Drushel, Ind. Eng. Chem.
Process Des. Dev., 5 (1966) 193.
R-W. Blue and C.J. Engle, Ind. Eng. Chem., 43 (1951) 494.
J.B. Butt, Advan. Chem. Ser., 109 (1972) 259.
E.E. Wolf and F. Alfani, Catal. Rev;Sci. Eng., 24 (1982) 329.
P. Forzatti, G.B. Ferraris, M. Morbidelli and S. Carra, Int. Chem. Eng., 24
(1984) 60.
8. Delmon, Appl. Catal., 15 (1985) 1.
D.L. Trimm, Appl. Catal., 5 (1983) 263.
B.W. Wojciecho~ki, Can. J. Chem. Eng., 46 (1968) 48.
K.J. Klingman and H.H. Lee, AIChE, 32 (1986) 309.
53

28 E.E. Wolf and E.E. Peterson, J. Catal., 47 (1977) 28.


2g J. Corella and J.M. Asua, Ind. Eng. Chem. Process Des. DeV., 21 (1982) 55.
30 0. Levenspiel, J. Catal., 25 (1972) 265.
31 S.S. Bhavikatti and S.R.Patwardhan, Ind. Eng. Chem. Prod. Res. Oev., 20
(1981) 106.
32 F. Yoshida, 0. Ramaswamy and C.A. Hougen, AIChE J 8 (1962) 5.
33 N.R. Draper and H. Smith, Applied Regression Analyiis, John Wiley & Sons
Inc., New York, (1966) 458.
34 J.L. Kuester and J.H. Mize, Optimization Techniques with Fortran, McGraw-
Hill Book Co., New York, (1973) 203.

You might also like