Pergamon

PII:

Chemical Engineerin9 Science, Vol. 53, No. 1, pp. 149 155, 1998 I 1997 Elsevier Science Ltd. All rights reserved Printed in Great Britain S0009-2509(97)00282-0 ooo9 2509/98 $19.00 + 0.00

Modelling of fixed-bed reactor: two models of industrial reactor for selective hydrogenation of acetylene
M. Szukiewicz, K. Kaczmarski and R. Petrus*
Department of Chemical Engineering and Process Control, Rzesz6w University of Technology, A1 Powstafic6w Warszawy 6, 35-959 Rzeszdw, Poland (Received 5 August 1996; accepted in revised form 2 May 1997) Abstract--Modelling and simulation of operation conditions for a heterogeneous fixed-bed reactor are investigated on the basis of the industrial reactor for selective hydrogenation of acetylene on palladium catalyst (in ethylene production). Heterogeneous and pseudohomogeneous models of the above-mentioned reactor with their advantages and drawbacks are presented. It is shown, that the proper formulation of a pseudohomogeneous model allows for its use even in process with complex reaction routes and in which, intraparticle mass transfer resistances plays a major role. 1997 Elsevier Science Ltd Keywords: Heterogeneous catalysis; selective hydrogenation of acetylene; simulation.

INTRODUCTION

Modelling and simulation have become more and more the usually applied methods of design and operation condition analysis of chemical apparatus. Unfortunately, in many chemical engineering branches using such convenient methods is difficult, especially sophisticated are cases, when we try to describe a real industrial process. If, additionally, the solid phase takes part in the process, e.g. heterogeneous catalysis, the task becomes very complex. The models commonly applied in simulation of heterogeneous fixed-bed reactor performance are widely described in literature (e.g. Froment and Bischoff, 1979; Doraiswamy and Sharma, 1984; Szarawara et al., 1991), but often there appears a problem of choice between two types of models: heterogeneous and pseudohomogeneous ones. The first type of model usually gives results with quite good accuracy, but solving the model equation set is rather difficult, because of a large set of highly non-linear and coupled equations (the usual case in industrial practice). Apart from that, even finding a solution method does not mean fully successful simulation since the time of solving can be unacceptably long. Therefore such models can be used only in a limited range, i.e. for preliminary calculation of reactor operation conditions. Solving pseudohomogeneous model is simple, but most often their accuracy is low, which limits their application. Therefore, it seems purposeful to make such a model, which

* Corresponding author.

would have the combined advantages of the two mentioned models it should be as precise as a heterogeneous one and as easy for solving and fast as a pseudohomogeneous one. The selective hydrogenation of acetylene on palladium catalyst (in ethylene production) is an example of a process, where intraparticle mass transfer resistances play a major role. Because of that, a prognosis for using typical pseudohomogeneous models is rather poor. The confirmation of this fact one can find even in the recently published paper by Schbib et al. (1996). In spite of good agreement between computed and experimental results in case of kinetic equations, the simulation of the behaviour of a commercial hydrogenation unit rather failed. In this work two models of industrial fixed-bed reactor for the selective hydrogenation of acetylene are presented. Both models were used for successful simulation of performance of industrial fixed-bed reactor for selective hydrogenation of acetylene (Szukiewicz et al., 1994; Szukiewicz, 1995). The process kinetics and reactor operation conditions were proposed by Fertilizers Research Institute, Pulawy, Poland. The first proposed model--the heterogeneous one--is very universal. Calculations based on it is acceptable for a wide range of changeable process operation conditions, but it has a great d r a w b a c k - - a quite long time of computations. The second--the pseudohomogeneous model is significantly faster than the heterogeneous one, at the cost of time necessary for carrying out the preliminary calculations and universality (it can be used only for such catalyst pellet for which preliminary computations were 149

150

M. Szukiewicz et al.

performed). These models can be used even for acomplex reaction set with sophisticated kinetic.
MODELS

Heterogeneous model Mass and heat balances for fluid: dFi 3(1 - -4-- - dV Ro
~) ko. i

Assumptions for both the models given below are the same: 1. Inside the tubular reactor, on catalyst pellet two exothermic reactions take place:
C2H 2 + H 2 ~ C2H4 C2H4 + H2 -~ C2H6.

RgT

(Pi,b -- Pi,s) = 0

i=A,B,C,D

{6)

]
dV

cpkldV

+ dV { - A H 2 ) ]

(7)

(1) (2)

with initial conditions for V = 0

The kinetic of these reactions are expressed in the following form: kl e x p ( - E / R g T)pA PB rc~n~ = 1 + K PA (3)
k2 FC2H6

PA = PA.O; PB = PB,0; PC = PC,O; PD = Po,o; T = To.

Mass balances for catalyst pellet:
Def,i R21

(8)

exp ( - E/Rg T ) PB + k~ exp (~-~E~~ pn. (4) ~:g ~ / 1 +KpA

RgT

~0 11~2~ ) -- a{R)r~ = 0 ~Pi'~

(9)

2. Activity distribution function can be described as follows: {~dla 0 ~ < R < R , dla R, ~< R ~< Ro.

i=A,B,C,D
where:
rA
=

rC2H2

a(R) =

r B = rC2H2 + rC2H6

(active material in outer part of catalyst pellet only)

r C = --rC2H2 -/- FC2H6

(lO)

(5)
3. The flow is of a plug type. 4. Axial mixing is ignored. 5. External and internal heat transfer resistances are negligible.
r D = _ rC2H~

with boundary conditions ~?pi --=0 dR for R = 0 (lla)

Remarks Kinetic expressions are proposed by Fertilizers Research Institute, Putawy, Poland (manufacturer of industrial catalyst) so as to fit in the best way to the experimental results. In our opinion, the success of a simulation of reactor operation conditions depends on the 'quality' of the kinetic equations. They, first of all, should fit as well as possible the experimental results and possibly have a simple form. Therefore, the kinetic equations proposed above should be regarded rather as empirical correlations which predict the rates of reactions within the range of experimental conditions. The complex mechanism of the selective hydrogenation of acetylene process, still partially speculative, induces that the equations presented here are as good as others, derived by using different possible reaction schemes. Assumptions 4 and 5 were accepted after checking them by using the well-known criteria (Fogler, 1992; Doraiswamy and Sharma, 1984) adapted to investigate the process (described in detail in Szukiewicz, 1995).

~3pi Def.i~--~ = kg,i(Pi,o - P i . s )

for R = Ro. (llb)

The model equations were solved using the Euler method for reactor mass and heat balances (6), (7) and orthogonal collocation method for pellet mass balances (9).

Pseudohomogeneous model Mass and heat balances:

dF~ d---~ + r* = 0 dT dV 1 cp

i=A,B,C,D

(12) (13)

(-AH1) +

d,o (--AH2) ]
dV

with initial conditions for V = 0 PA = PA,O; Pn = PB.O; PC = PC.O;

Po=PD,o;

T=To

(14)

Modelling of fixed-bed reactor

151

where
r* = rll rC2H2,b

r~ = ~1 rc:H:,b + ?/2 rC2H~,b rE = - q l rc2H2,b + r/2 rC2H~.b

r* = --q2 rC2H~,b.

(15)

Effectiveness factors were defined for both main reactions: ql - ~v'rc2n2dVp

Vp FC2Hz,b

(16a)

r12 - ~vf C:H~d Vp

Vp FC2H~,,b

(16b)

In order to minimize time of calculation with the use of this model, the values of the effectiveness factors in eqs (15) were computed on the basis of the following scheme: it was assumed that the functional dependence of effectiveness factors has a form qj = f j ( r , yA,y,) j = 1, 2. (17)

is most often very long. It is the largest drawback of the heterogeneous model presented above; therefore, the pseudohomogeneous model was developed. As it will be shown in the following, the pseudohomogeneous model proposed here is as precise as the heterogeneous one, but significantly easier for solving and many times faster. In this case, the main problem is the proper assumption about the functional dependence of the effectiveness factors. This assumption decides about good or poor effect of simulation. A next step (problem) is the base building. It should ensure an acceptable precision of calculations, which means finding the proper way of discretization of space (T, YA, YB). Of course, the more points in the base the better, because approximation of functions eq. (17) is more precise, but it also means more time lost on preliminary calculations of the effectiveness factor values. For the process under consideration, the proper discretization of space (T, YA, YB) w a s found by trial-and-error method; a dimension (amount of points) of the base was enlarged until differences between the computed results using both models were not more than 1% (for both the criteria described below) for the chosen reactor operation conditions. A comparison of the agreement between the two proposed models were done on the basis of the following parameters: Acetylene conversion:
~a --

For chosen points of space (T, YA, YB) effectiveness factor values were calculated; these points were chosen in such a way as to contain each possible operation conditions of the modelling reactor. Such set of points together with the effectiveness factors values calculated for them will be called, 'the base' in the following. The calculations of the effectiveness factor values were performed using the model of a single catalyst pellet eq. (9). Expanding the function (17) in Taylor series (limited to the first derivatives only) qj(T + AT, y~, + AyA, y~ + Ayn) = tlj(T, ya, y~) c3~IJ(T"ya'YB)A T + &lj(T, YA, Yn) AyA -I OT ~YA -t ~qj(T, yA,yB) OYB and using an interpolation method, it is possible to calculate the effectiveness factor values in eqs (15). The model equations (12) and (13) were solved using the Euler method.
RESULTS AND DISCUSSION

PA.O -- PA
-

(19)

Pmo

Selectivity: S = (hydrogen consumption in acetylene hydrogenation reaction/total hydrogen consumption).

(20)
Comparison of the results of simulation, in a wide range of reactor operation conditions are presented in Figs 1-4. Tables 1 and 2 present the relative errors between the computed results using both models for the chosen reactor operation conditions. All of the figures and tables show that compatibility of the results obtained for both models is very high. For many cases results of the simulations are exactly the same (curves on figures overlaps at least partially) or differences are significantlysmaller than the assumed 1% of relative error (Tables 1 and 2). This error was exceeded only near the higher limit of the temperature range of reactor operation conditions (Figs 3 and 4, Table 2), which was accepted as tolerable. It could be corrected by changing the base discretization (which practically means that the base will grow) but at the cost of computational time. Higher differences for higher temperatures, in our opinion, come from the fact that an influence of temperature is strongly nonlinear, while method used for the calculation of effectiveness factor is simply a multidimensional linear approximation. Comparison of Figs 1 and 3 with Figs 2 and 4, respectively, shows that compatibility

Ays,

j = 1,2

(18)

Heterogeneous models of fixed-bed reactors give results of quite good precision, but especially in case of complex reaction with many reaction components and sophisticated kinetics their solution is very difficult. Finding a method of solution is a very difficult task (see Szukiewicz, 1995). Apart from that, the time of solution of the heterogeneous model equations set

152
1.0

M. S z u k i e w i c z et al.

Eo=1.4

HET

'7
<

0,8

] ...... PSH

t
0.6

E0=1.2

E .o
ID ;> O O ID ,..) 4.0 O

= .

0.4

0.2

0,0
0.0 0.2 0.4 0.6 0.8 1.0

r e a c t o r lenght, z[-] Fig. 1. Calculated conversion for heterogeneous and pseudohomogeneous models for chosen values of parameter E0 = Ps,o/PA,o.

1.0

. ~ =

- .

E0=1.1

"7"
c#3

0.9

E=1'4

" ;'"

- "

"3

- o,3 0.8

HET PSH

" " ~ ,

0.0

0.2

0.4

0.6

0.8

.0

r e a c t o r lenght, z[-] Fig. 2. Calculated selectivity for heterogeneous and pseudohomogeneous models for chosen values of parameter E0 = PB,o/PA,O.

between the computed results in the case of acetylene conversion is much better than in the case of selectivity. Precision of the pseudohomogeneous model for conversion is practically the same as the heterogeneous one. Accuracy of selectivity calculation is in most cases worse. In the authors' opinion, the difference in precision of calculation for conversion and selectivity values is caused by the fact that the first parameter depends on acetylene hydrogenation reaction only, while the second depends on both the reactions considered here. As it was shown, results of calculation using both models are much the same, but differences can be

larger or smaller in dependence on the way of discretization, the interpolating method used and, finally, the calculated parameter. The difference can be reduced by enlarging the base dimension, but at the cost of computational time. The evaluation, whether accuracy of the obtained results (for determined process operation conditions) is sufficient or not, is up to the model builder. Times of calculation using the heterogeneous and the pseudohomogeneous models and times needed for the base preparation and simulation are presented in Table 3. They show that calculations using the second model are more than 10,000 times faster.

Modelling of fixed-bed reactor 1.0

153

"7"
<

0.8

HET

"'

. . . . . . . . . . . . . .

------

o o,,,~
cD

0.6

o o

0.4

I1) O

0.2

0.0

290

300

310

320

330

340

front-end temperature, T0[K ]

Fig. 3. Calculated conversion for heterogeneous and pseudohomogeneous models; Eo = 1.2. 1.0

- ~ ' ' - . - .....

HET
PSH

,-p r/l

0.9

0.8

290

300

310

320

330

340

front-end temperature, T0[K ]

Fig. 4. Calculated selectivityfor heterogeneous and pseudohomogeneous models; Eo = 1.2.

Unfortunately, this model also needs a time-consuming preliminary calculations and it is less universal (it can be used only for such catalyst pellet, for which the preliminary computations were performed). These calculations need the model, which can simulate behaviour of a single catalyst pellet for the investigating process. Our investigations on modelling of the selective hydrogenation of acetylene process allows for listing the following drawbacks and advantages of both the models presented: The heterogeneous model: Advantages--accuracy of computations is very high, the model is very universal, which

means that it is possible to change practically all parameters without rebuilding of the model. Drawbacks difficulties in solving model equations set and a long time of calculations. The pseudohomogeneous model: Advantages--very short time of computation; therefore, this model can be used for optimization of reactor operation conditions and even for online control. Drawbacks--an acceptable precision of calculations needs a large size base, it is very time consuming, apart from being less universal; the pseudohomogeneous model can be used only for

154

M. Szukiewicz et al. such type of catalyst pellet for which the prelimin a r y c o m p u t a t i o n s were performed. any c o m p u t a t i o n s of o p t i m a l operation conditions would be practically impossible (even when h a v i n g m a n y times faster calculating c o m p u t e r t h a n ours) by using the heterogeneous model only. M a t h e m a t ical optimization needs m a n y repeated simulations of the investigated unit o p e r a t i o n conditions, a n d only a fast model ensures its usefulness. In the investigated process, optimization were perform e d on a c o m p u t e r which is not very fast I B M PC 486DX-33 MHz.

Because of the a d v a n t a g e s a n d d r a w b a c k s m e n t i o n e d above, the presented models were used for hydrogenation unit simulation s u p p l e m e n t a r y - - t h e heterogeneous one for preliminary calculations, a n d the p s e u d o h o m o g e n e o u s one for finding of optimal operation conditions. It is worthwhile to a d d that

O n the basis of the investigations performed for selective h y d r o g e n a t i o n of acetylene process some 2 ~A,IHET) 0~A,(PSH) ~ (%) S(HET) S(PSH) ~ (%) n o t e w o r t h y conclusions regarding heterogeneous 0.955 0.5 fixed-bed reactor modelling can be drawn. Each of the 0 0.000 0.000 0.960 two models presented has its o w n d r a w b a c k s a n d 0.1 0.065 0.065 0 0.958 0.954 0.4 0.952 0.4 advantages a n d therefore it is n o t possible to present 0.2 0.135 0.135 0 0.956 0.949 0.4 a universal m e t h o d o l o g y which will be correct in 0.3 0.211 0.211 0 0.953 0.945 0.5 every case. We are able to ,present only some direc0.4 0.291 0.291 0 0.950 0.942 0.5 tions, which in o u r o p i n i o n can help to do successful 0.5 0.371 0.371 0 0.947 0.6 0.452 0.452 0 0.943 0.937 0.6 simulation of o p e r a t i o n of any industrial hetero0.7 0.536 0.536 0 0.938 0.930 0.9 geneous reactor. 0.8 0.619 0.619 0 0.931 0.923 0.9 The heterogeneous model is very convenient for 0.9 0.701 0.701 0 0.924 0.916 0.9 investigations in case of a wide range of process para1 0.780 0.780 0 0.915 0.908 0.8 meters changes, but finding its solution is usually a rather difficult task. The p s e u d o h o m o g e n e o u s Relative errors were calculated as 6 = (~a.(nET) -- eaaPSH))/ model suggested in this p a p e r is easy to solve a n d very ~A,(HET)or 6 = (S(HET)-- S(PSH))/S(HET ), fast, but at the cost of time necessary for carrying out the preliminary calculations a n d less universality (universality of the model depends on the investigated Table 2. Relative errors for acetylene conversion and proprocess). cess selectivity for heterogeneous and pseudohomogeneous In o u r opinion, in m a n y processes (e.g., the selective models for chosen front-end temperatures, Eo = 1.2 h y d r o g e n a t i o n of acetylene process presented here) the models m e n t i o n e d above are s u p p l e m e n t a r y a n d it Z ~A,(HET) ~A,(PSH) 6 (%) S(HET ) S(PSH ~ (%) ) is purposeful to build up b o t h of them: First, a hetero0.951 0.3 geneous one to approximately determine reactor 293 0.190 0.190 0 0.954 303 0.434 0.434 0 0.943 0.938 0.5 o p e r a t i o n conditions and, then, the p s e u d o h o m o 313 0.778 0.773 0.6 0.905 0.898 0.8 geneous to optimize these conditions. P r e p a r a t i o n of 323 0.964 0.96 0.4 0.820 0.821 -0.1 the heterogeneous model of reactor first means, in 333 0.960 0.950 1.0 0.802 0.795 0.9 practice, that a model for calculating the effectiveness 343 0.941 0.920 2.2 0.786 0.765 2.7 factor (model of a single catalyst pellet) is ready, too. This strategy is correct, especially in case when Relative errors were calculated as 6 = (ea.~UEV)-- C~A.~PSH))/ universality of b o t h models is n o t the same. If ~A,(HET)or 6 = (S(HET)-- S(PSH))/S(HET ). a p s e u d o h o m o g e n e o u s model is as universal as the heterogeneous one or the time needed for p r o p e r base calculation is significantly shorter t h a n the time of Table 3. The comparison of times of calculation a single reactor work simulation using the heterogeneous model, it seems t h a t it is better to build it at once. Pseudohomogeneous model The model of a single catalyst pellet is significantly The calculation of r/1 and r/2 an average time 8 min easier for solving t h a n the heterogeneous model of using single catalyst pellet model 20 s reactor and, therefore, calculating even a large size Preliminary calculations of the base total time about 78 h base is better, because we o b t a i n immediately a very 30 min A single reactor work simulation an average time 1 s useful calculating tool; a very fast a n d precise model can be used, e.g. for process o p e r a t i o n optimization or Heterogeneous model on-line control. A single reactor work simulation an average time 6 h The influence of different (especially larger) n u m b e r 20 rain of reaction c o m p o n e n t s a n d / o r the n u m b e r of reacThe results presented in the table were performed using tions, which is to say ' a n o t h e r process', o n the time of computer IBM PC 486DX-33 MHz. c o m p u t a t i o n using b o t h the heterogeneous a n d the

Table 1. Relative errors for acetylene conversion and process selectivity for heterogeneous and pseudohomogeneous models for chosen points along the reactor, Eo = 1.2

CONCLUSIONS

Modelling of fixed-bed reactor pseudohomogeneous model is difficult to precisely foresee, when details of the process are unknown. Of course, apart from the simplest cases (e.g., one substrate, linear kinetic, etc.) where the profit of using the pseudohomogeneous model will be too small to build such a model. In our opinion, the final evaluation of computational expense of both models or only one of them is up to the model builder,
NOTATION

155

Ayi z = 1/L

change of mole fraction of component i dimensionless distance along the reactor

Greek letters ~A 6 e, th, q2 Subscripts 0 1 2 b p s A B C D

conversion of acetylene relative error porosity effectiveness factors

activity profile function heat capacity of reacting gas Cp effective diffusion coefficient of Def,i component i activation energy E Eo = PB,o/PA,o front-end hydrogen-acetylene ratio molar flow rate of component i Fi heat of reaction AH1, AH2 reaction rate constants kl, k2, k~ mass transfer coefficient of comk0,i ponent i K adsorption equilibrium coefficient l position in reactor L length of reactor partial pressure of component i Pi P total pressure reaction rate of component i for ri heterogeneous model r* reaction rate of component i for pseudohomogeneous model reaction rate defined by eq. (3) rCzHz reaction rate defined by eq. (4) rCzH6 R radial position in catalyst particle Ra radius of pellet on board inactive core-active shell gas constant Ro radius of catalyst pellet Ro S selectivity T temperature AT change of temperature V reactor volume Vp catalyst pellet volume mole fraction of component i Yi

a(R)

reactor inlet hydrogenation of acetylene reaction hydrogenation of ethylene reaction bulk pellet pellet surface acetylene hydrogen ethylene ethane
REFERENCES

Doraiswamy, L. K. and Sharma, M. M. (1984) Heterogeneous Reactions: Analysis, Examples and Reactor Design. Wiley, New York. Fogler, H. (1992) Elements of Chemical Reaction Engineering, Prentice-Hall International, Inc. Englewood Cliffs, NJ. Froment, G. F. and Bischoff, K. B. (1979) Chemical Reactor Analysis and Design. Wiley, New York. Schbib, N. S., Garcia, M. A., Gigola, C. A. and Errazu, A. F. (1996) Kinetics of front-end acetylene hydrogenation in ethylene production. Ind. Engng Chem. Res. 35, 1496 1505. Szarawara, J., Skrzypek J. and Gawdzik, A. (1991) Podstawy In~ynierii Reaktorbw Chemicznych. WNT. Szukiewicz, M. (1995) Ph. D. thesis. Cracow Technical University, Poland. Szukiewicz, M., Kaczmarski, K., Petrus, R. and Got~biowski, A. (1994) Selective hydrogenation of acetylene. Mathematical analysis. In~. Chem. I Procesowa 3, 481-489.