- Email: [email protected]
Behavior of a pilot plant fixed-bed reactor during catalyst deactivation
Chemical Printed
Engineering in Great
Science,
Vol. 41, No. 4, pp. 779-786,
1986.
0
Britain.
BEHAVIOR
OF
A
PILOT
A.
Swiss *Institute
PLANT
Baiker*.
*Department Federal Institute of Physical
Dedicated
to
FIXED-BED
Professor
0.
REACTOR
DURING
and
Epple*.
A.
CATALYST
Wokaun
000%2509/86 $3.00 + 0.00 1986. Pergamon Press Ltd.
DEACTIVATION
**
of
Industrial and Engineering Chemistry, of Technology (ETH). CH-8092 Zurich, Switzerland Chemistry. University of Bayreuth, D-8580 Bayreuth
W.
Richarz
on
the
occasion
of
his
60th
birthday
A%STRACT the hydrogenation of non-isothermal fixed-bed reactor used for The behavior of an non-adiabatic, addihas been studied under the influence of catalyst deactivation caused by continuous toluene profiles Axial and radial temperature and concentration of thiophene to the reactant feed. tion (Biot In spite of significant radial temperature gradients measured during deactivation. were pseudoit was possible to describe the reactor behavior with a one-dimensional = 5.25), number proMean cross-sectional temperatures can be used since the measured radial homogeneous model. The activation energy for thiophene adsorption derived from the measured parabol i c. files are profiles (4.7 kJ/mole) agrees with previous direct determinations in the literature.
KEYWORDS Fixed-bed and axial
catalytic temperature
toluene reactor; and concentration
INTRODUCT
hydrogenation: profiles:
catalyst deactivation: feed mean cross-sectional temperature:
poisoning: reactor
radial model _
ION
Although the behavior of fixed-bed reactors during catalyst deactivation has received considerable in particular because of its importance in industry, there attention (Butt, 1980: Hughes, 1984). pro. relatively little experimental work where axial and radial temperature and concentration Pexidr. Cerny and Pasek (1968) measured have been measured duri ng catalyst deactivation. +fles axial temperature and concentration profiles for the benzene hydrogenation on a Ni catalyst during Weng. Eigenberger and Butt (1975) studied the deactivation of a non-isothercli;ctivation by CS2. feed. non-adiabatic fixed-bed reactor for the same reaction during thiophene addition to the for this reason no radial temperature The employed reactor had a small inner diameter (0.8 cm): Price and Butt (1977) studied the factors which play a decisive role for an a gradient existed. Richardson (1971) and Wheeler and priori simulation of the reactor behavior during deactivation. parameRobe11 (1969) showed that for isothermal reactors with constant overall deactivation rate ters the activity profiles could be described by the adsorption theory of Bohart and Adam (1920). In the present work we have measured axial and radial temperature and concentration profiles in a An efficient modelling of the observed reactor during deactivation by feed poisoning. fixed-bed mean was achieved by means of a simple pseudo-homogeneous one-dimensional model using a profiles The basic concept is to estimate kinetic and heat transfer parameters cross-sectional temperature. and to assume that the only influence of thiophene adsorption is to from profiles. stationary Thus the deactivation rate parameter can be deterreduce the active surface area of the catalyst. mined by evaluating the profiles taken at a series of times after initiation of deactivation.
EXPERIMENTAL Fig. 1). It The fixed-bed reactor pilot plant has been described by Baiker and Bergougnan (1985a. A jacketed reactor tube of 2.2 m wa* operated as a continuous reactor with recycle of hydrogen. A diagram of the reactor showing the positions of length and 0.052 m inner diameter was employed. (1986. Fig. 1). The catalyst bed and measuring devices has been given by Baiker and Epple the catalyst bed of 0.7 m length was embedded in inert glass packings on both sides. Axial temperature separated by seven positions in the catalyst bed and concentration profiles were measured at 1 and 2 0.1 m ( position 3: z = 0, catalyst bed entrance: position 9: z = 0.6 m: positions The radial temperature and were located in the inert glass bead packing at the reactor entrance). profiles were measured at five radial positions at z = 0.25 m and z = 0.55 m. concentration infrared wall temperature was controlled by circulating oil through the jacket. An The reactor
779
A.
780
BAIKER
D-5
etai.
anaiyzer (URAS) was utilized for the automatic analysis of the toluene concentrations. The plant acquisition were controlled by a PDP ll/lO process A commercial operation and data computer. Ni catalyst (ICI 42.1 ) was used. Properties of the cylindrical (3.4 x 3.4 mm) silica supported pellets were given elsewhere (Baiker and Epple, 11986). Before use the catalyst was catalyst reduced according to the procedure given by the manufacturer. Toluene (99.5 X). nitrogen (99.99%). without further purification. and hydrogen (99.99%) were used Recording of the axial and radial profiles was started after steady-state conditions were reached in the reactor, typically after less than 20 minutes. Then poisoning by addition of thiophene to the reactor feed was initiated. Results of four deactivation experiments performed at 2 bar are presented in this paper. They were characterized by the following conditions. Run (A):Fte;pe;;; ture of oil in reactor jacket. TB = 381 K: total flow rate (hydrogen and toluene). moles/h: Reynolds number, Re = 99: 2.4 vol-%; toluene concentration. thiophene feed, 12 ml/h; total. &a: TB= 393 K; F = 850 moles/h: Re = 104: toluene concentration, 1.8 ~01%; thiophene feed. 25 ml/h. TRE, = o:392 K; toluene concentration, 1.5 ~01%: $nmtFi: F = 690 moles/h, . Re = 78: TB = 394 K: F = 640 moles/h: thiophene feed. Re = 100: toluene concentration. 2.8 vol%: feed. 26 ml/h. thiophene The thiophene adsorption capacity of the catalyst at full deactivation was measured by means of the method described by Granatelli (1959). in which the amount of sulfur is determined. A mean sulfur content of 2.6 mg sulfur per g of catalyst was found. This corresponds to a thiophene adsorption capacity of 0.08 moles per g of catalyst.
MEASURED
AXIAL
PROFILES
Figure 1 depicts the time profiles of temperatures and toluene concentrations and Epple. axial positions 3 - 9 (Fig. 1, Baiker 1986) during deactivation four-point interpolation procedure used to derive continuous axial and radial measured data has been described by Baiker and Bergougnan (1985a).
150
100
Time
Fig.
1
Measured time profiles at positions 3 - 9.
Run
of axial (A).
measured experiment profiles
200
(min)
temperatures
and
toluene
concentrations
at (A). from
the The the
781
Catalyst deactivation of a fixed-bed reactor
D-S
Note that in the initial period Figure 2 shows the axial profiles of experiments (A) and (8). deactivation (t = 5 min) the hot spot temperature was significantly higher than the one of the hot spot shifted towards the With progressing deactivation. steady-state profiles (t = 0). decreased. and the hot spot temperature the profiles were broadened, of the catalyst bed, an inactive zone at the entrance corresponding axial toluene concentration profiles indicated conversion decreased to about which is formed relatively late, i. e. after the catalyst bed, of its initial value.
Catalyst
bed 2
packing
Inert
Temperature
+
bed
Catalyst
I
1
I
Temperature -
1
1
profiles
I
Fig.
2
I 0.7
0.0 axial
I
I
coordinate
z (ml
Interpolated axial temperature and toluene concentration profiles before of runs (A) and (B). Curve 1 corresponds to steady-state deactivation. Curves 2 - 9 represent reactor profiles during deactivation measured in intervals of five minutes after initiation ranges: temperature. of thiophene addition to the feed. Variable 0 - 250 OC: toluene concentration, 0 - 3 ~01%.
of the end The of 20%
A.
782 MEASURED
RADIAL
BATKER
et al.
D-5
PROFILES
The radial temperature profiles measured during deactivation experiments (Fig. 3) were similar in shape to those found during steady-state operation. for which a Biot number h R / k = 5.25 was As a consequence of the parabolic shape found in all experiments.wmean c%oss-sectiocalculated. temperatures could be defined and applied for the reactor model. nal The procedure used for the estimation of the mean cross-sectional temperature has been described by Baiker and Epple (1986). Measured radial toluene concentration profiles were f1at.i.e. the mean cross section concentrations corresponded approximately to the concentrations in the center of the reactor tube.
100
C
100
MINUTES . cl . V m .
IC
1
0 4 8 12 10 28
MINUTES V 0
6 12
L
= 0.55
M
1 (r/R)
100 ID
(r/R)
v 0 l
v A = *
I
I
MINUTES 0 10 12 14 16 18 26
z = 0.25
D
MINUTES 7
10
0
8
m
0
M
1 (r/R)
Fig.
3
Typical radial z - 0.25 m and (Runs C and D;
1 (r/R)
temperature profiles measured at axial positions z = 0.55 m during deactivation experiments. TB = temperature of oil in reactor jacket)
D-5
Catalyst REACTOR
deactivation
of a fixed-bed
reactor
783
MODEL
The reactor model is based on the assumption that the kinetic parameters of toluene hydrogenation thus the kinetic equations for toluene hydrogenation are not affected by the deactivation process: and for the catalyst deactivation are separable. Consequently the heat transfer parameter and the rate constant could be estimated from the initial steady-state of the reactor. In hydrogenation order the deactivation behavior. the model was extended by introducing a linear to describe equation differential equation for the poisoning of the active surface area, and a mass balance for thiophene. The one-dimensional vation is given Heat
pseudo-homogeneous the following set
by
reactor model equations.
of
E
u
Pf
2 lJ (Tb
aT
Cpf
-
T)
+ at
(l-s)Ps
ac
CPS az
R
Boundary
ac
(1-E)
a2
c
the
behavior
without
(l-s)~s
c
Ps
AH r -rr(T.
PT)
deacti-
(1)
=Ps
conditions:
Initial
conditions:
The kinetic investigation
rate (Baiker
r(T.PT)
T(z
= 0.
t)
= Tin(t)
C(z
= 0.
t.)
=
T(z.
t
= 0)
= Tin(t
C(z.
t
= 0)
= 0
= 0)
the toluene 1985b).
(P,/FT)
of catalyst
expIK8
the reactor was defined
(l/T
-
=
pthio
The rate parameter mean temperature T between the estimatzd The second relationship
thio’dt
-
-
been
discussed
in
adsorbed
at
adsorbed
at
total
the
deactivation,
time-dependent
e(t)
k(T,)
k(T) = k according parameters required
u dC thio/da
+
[(l-
exp[-Es/R
(l/T
-
~1 Ps
MT
active
frac-
(4)
Kul’kova (1962) This suggests
have that
found a linear dependence of the time-dependence of g(t)
l/Tm)l.
exp[-E /(RT)], of to theaprocedure of k(T ) and E . to !escribeadeactivation
c)/
earlier
t deactivation
and e(t).
an
(3)
during equation
Andreeva Richardson (1971) and Lyubarskii. the deactivation rate on the active fraction is governed by an equation of the form
/ dt
has
1 thiophene
-d e(t)
hydrogenation
l/Kg))
behavior by the
thiophene =
Cin(t>
=
description of the
e(t)
(2)
r(TvPT)
equation used for and Bergougnan.
1 f
the O(t)
Ps
u--
-
at
dC
describe
bal ante:
m=
For tion
to
bal ante: aT
Mass
used
dimensions Himmelb’lau
d g(t)/dt.
is
bar-’ (1970). the
s-‘,
to
thiophene
has been reduce mass
transformed to the correlation
balance,
a
A.
BAIKER
used
(t
784 initial to the
The following adding thiophene
conditions feed):
were
T(z.
t
=
0)
=
T(z)
(temperature
C(z,
t
=
0)
=
C(z)
(toluene
t
‘thio(” O(z.
t
=
=
0)
Furthermore
0) =
the
=
0
boundary
conditions
=
0.
t)
=
Tin
(temperature
C(z
=
0.
t.)
=
Gin(t)
(toluene
Numerical
=
o.t)
=
(t)
Cthio
is
the
state in
time
before steady-state
when
deactivation
is
started
by
deactivation) before
deactivation
activity).
T(z
C thio(z
steady
0
D-S
concentration)
(initial
following
=
concentration
(thiophene
1
in
et al.
(thiophene
were at
applied:
entrance
concentration
to in
concentration
catalyst
bed)
feed) in
feed)
Solution
orthogonal collocation (Villadsen and Michelsen. The method of 1978) was used to reduce the partial differential equations to a set of ordinary differential equations. In the present work 8 The resulting system consists of 24 differential collocation points were equations, 8 used. algebraic equations for the deactivation, and the deactivation equation at bed entrance (z = 0). This system of equations was simultaneously integrated using Gear’s variable i ntegrati on step These integration routines were used together algorithm, as provided in the software package IMSL. with the nonlinear multiresponse regression programm RKPES (Klaus and Rippin. 1979). This program is based on the “maximum 1 i kel i hood” method (Bard, 1974). RKPES was used with the option of an The experimentally measured data were interpolated with unknown non-diagonal covariance matrix. since the positions of respect to time and space previous to use in the regression program, the measurements did not correspond to the positions of the collocation points, simultaneous and a recording of all measurement points was not possible. Estimated
Parameters
from
Steady-State
Profiles
The kinetic experiments 8.8 -I_ 1.1 to depend
parameters and the overall heat transfer coefficient were estimated from steady state and reported elsewhere (Baiker and Epple. 1986). Mean parameter values found were: k,. mole /kgs: 4.25 + 0.03 K. The heat transfer parameter U was found 7.3 + 1.8 K: k3, on the Reynol kg’s number (cf. Fig. 4 in BTiker and Epple. 1986).
Estimated
Deactivation
Parameters
Deactivation parameters k(T ) and procedure described above. Thg data 30. and 40 minutes CfteT initiation ; E /R. 530 k(T >, 1.05 r. 0.08 s whi!!!h is due to the small $bsolute kJ/mole indicates a weak temperature
were estimated from eight deactivation experiments E by the s& used consisted of the profiles determined at times 10. 20, of the deactivation process. Mean parameter values found were: + 145 K. Note the relatively 1 arge standard deviation of E /R, value of the activation energy. The activation energy of a4. 7 dependence of the deactivation.
DISCUSSION Calculated
and
Experimental
Profiles
Experimental and calculated profiles of runs (A) and (B) are compared in Fig. 4The model provides a good representation of the measured temperature and concentration profiles. The deviation experimental and simulated axial concentration profiles in run (A) at the reactor between outlet can be ascribed to a systematic error in the measurement of very low toluene concentrations. Deactivation
kinetics
The quality of the description of the reactor behavior showed a different sensitivity on the two It was very sensitive to the pre-exponential estimated deactivation parameters. factor k(T ). which in turn was found to be strongly correlated with the thiophene adsorption capacity. In c%the parameter E had only a minor influence on the quality of the fit. which indicates a trast. TRis activation energy of the deactivation kinetics, i.e. very shallow optimum. adsorption the rate of thiophene on Ni. Our value derived from the measured profiles is quite low (4.7 kJ/mole). agrees well with direct determinations in the literature. Eigenberger and Butt (1975) found Weng. while Zrncevic and Gomzi (1983) gave a value of 5.3 kJ/mole. a value of 4.5 kJ/mole. Considering the fact that the thiophene adsorption rate depends only weakly on temperature, an isothermal treatment of the deactivation reaction may lead to satisfactory results. No tion will
definite rate only
conclusions have and the deactivation lead to significant
been
reached concerning rate can be considered errors in the description
the to of
question whether the toluene be separable. However this the profiles when the catalyst
hydrogenaassumption activity
Catalyst
D-5
deactivation
of a fixed-bed
reactor
785
when the conversion in the reactor has dropped to values less than 20 % of the is low. i.e. initial conversion. In contrast to Weng. Eigenberger and Butt (1975) who used a semi-empirical two-s i te mechanism for the deactivation kinetics, we have employed a simple linear dependence between the toluene hydrogenation activity and the adsorbed amount of thiophene. This leads to a considerable reduction of computation time. Consequences of this simplifying assumption are noticeable only for high degrees of deactivation, which are generally of less importance in practical It appears that under the conditions given,the one-dimensional applications. pseudo-homogeneous reactor model can be considered as accurate enough for most practical purposes.
Experimental Temoerature
A
>alculated
orofiles
A
Temperature
profiles
?I
Experimental
Concentration
6
Calculated
pro
6
Concentration
profiles
I
;1
0
L Temperature
r
I
profiles
I
I
axial 4
’
I
’
I
,
coordinate
z
(m
profiles
I
0.7
0.c )
Fig.
perature
I
I
I
0.7
)
Experimental and calculated axial temperature and toluene concentration files of runs (A) and (B). Symbols in calculated curves represent T and values calculated at collocation points. Variable ranges: temperature, 100 - 250 OC: toluene concentration. 0 Curves 1. 2. 3. and 4 correspond to times of 10. 20. 30. and 40 minutes, after initiation of thiophene addition to the feed. respectivley.
pro-
C 3
~01%.
786
A.
D-5
BAIKER etal.
CONCLUSIONS The behavior of a fixed-bed reactor during deactivation has been investigated. Measured axi al prof i les were described by a temperature and concentration pseudo-homogeneous one-dimensional model. Descripton with a one-dimensional model was appropriate because the measured radial temperature profiles exhibited a parabolic shape during the entire course of deactivation. The use of a temperature is justified if the radial position corresponding to mean cross-sectional the mean This condition is fulfilled for radial profiles of the form cross-section temperRt.ure is constant. T(r) = T(0) + a(r/R) . A first order rate expression was found to be appropriate to describe the deactivation caused by thiophene adsorption on the supported Ni catalyst. The thiophene adsorption rate was found to be only weakly dependent on temperature, indicating that in several situations of practical i mportance the use of an isothermal deactivation kinetics may be sufficiently accurate. This simplification is very well justified for moderately deactivated catalyst beds. At high degrees of deactivation a more complex two-site model may be more appropriate.
LIST C r
Lthio cPs ZPf
Gp OH
Re
F
t
T U U z
E
9
3S
OF
SYMBOLS
-3 toluene concentration. mol m -3 thiophene concentration. mol m specific heat of catalyst, J kg-‘K-l specific heat of fluid phase, J kg-‘K-l catalyst particle diaye:fr, m mass flow rate, kg s m reaction enthalpy. J mol-’ heat transfer coefficient throygh,;etctor radial heat conductivity, J m s kinetic parameter, mol/kg kinetic parameter, K kinetic parameter, K -1 thiophene adsorption capacity. mol kg toluene partial pressure. bar mean toluene partial pressure. bar radius of reactor tube. m Reynolds number (G dp/u) radial coordinate, m _, -1 reaction rate, mol kg s time, s temperature, K J ,-zs-1 overall heat transfer coefficipnt. superficial gas velocity, m s axial coordinate, m bed porosity active fraction of catalyst syrface density of fluid phase. kg m-3 density of solid phase, kg m
wall,
J
m-2s-1
K--l
K-l
REFERENCES and M. Bergougnan (1985). Can. J. Chem. Eng.. 63. 138-145. Bai ker, A., and M. Bergougnan (1985). Can. J. Chem. Eng.. 63. 146-154. Bai ker. A., and D. Epple (1986). Appl. Catal., In press. Bai ker, A., Nonlinear Parameter Estimation. Academic Press, New York. Bard, Y. (1 974). and Adam, E.Q. (1920). J. Am. Chem. Sot., 42. 523-544. Bohart, G. S (~~8~~~5~atalyst Deactivation. Elsevier. Amsterdam. Butt, J.B. ) Anal. Chem., 31. 434-436. Granatelli, Himmelblau. D: M. (1950). Process Analysis by Statistical Methods. Wiley. NE!w York. (1984). Deactivation of Catalysts. Academic Press, New York. Hughes. R. and D. W. T. Rippin(1979). Computers & Chemical Engineering 3. Klaus. R.. 105-115. L. B. Andreeva, and N. V. Ku1 ‘kova (1962). Kin. Kat.: 3. Lyubarski i, 102-110. G. D.. Pexidr, V. . Pasek (1968). Proc. 4th European Symp. on Chs km. React .ion Eng.. I. Cerny. and I. Br,ussels. 239-251. Price, T.-H.. and-J. B. Butt (1977). Chem. Eng. Sci., 32. 393-412. Richardson. J.T. (1971). 3. Catal.. 1, 130-138. Michelsen (1978). Solution of Differential Equation Models Villadsen. J. V.. and M. L. Polynomial Approximation. Prentice Hall, New Jersey. G. Eigenberger. and J.B. Butt (1975). Chem. Eng. Sci.. 30. 1341-1351. Weng. H. S., Wheeler. A.. and A.J. Robe11 (1969). J. Catal.. 13. 299-305. Zrncevic. S.. and Z. Gomzi (1983). Chem. Eng. Sci.. 38. 1351-1354.
Engineering in Great
Science,
Vol. 41, No. 4, pp. 779-786,
1986.
0
Britain.
BEHAVIOR
OF
A
PILOT
A.
Swiss *Institute
PLANT
Baiker*.
*Department Federal Institute of Physical
Dedicated
to
FIXED-BED
Professor
0.
REACTOR
DURING
and
Epple*.
A.
CATALYST
Wokaun
000%2509/86 $3.00 + 0.00 1986. Pergamon Press Ltd.
DEACTIVATION
**
of
Industrial and Engineering Chemistry, of Technology (ETH). CH-8092 Zurich, Switzerland Chemistry. University of Bayreuth, D-8580 Bayreuth
W.
Richarz
on
the
occasion
of
his
60th
birthday
A%STRACT the hydrogenation of non-isothermal fixed-bed reactor used for The behavior of an non-adiabatic, addihas been studied under the influence of catalyst deactivation caused by continuous toluene profiles Axial and radial temperature and concentration of thiophene to the reactant feed. tion (Biot In spite of significant radial temperature gradients measured during deactivation. were pseudoit was possible to describe the reactor behavior with a one-dimensional = 5.25), number proMean cross-sectional temperatures can be used since the measured radial homogeneous model. The activation energy for thiophene adsorption derived from the measured parabol i c. files are profiles (4.7 kJ/mole) agrees with previous direct determinations in the literature.
KEYWORDS Fixed-bed and axial
catalytic temperature
toluene reactor; and concentration
INTRODUCT
hydrogenation: profiles:
catalyst deactivation: feed mean cross-sectional temperature:
poisoning: reactor
radial model _
ION
Although the behavior of fixed-bed reactors during catalyst deactivation has received considerable in particular because of its importance in industry, there attention (Butt, 1980: Hughes, 1984). pro. relatively little experimental work where axial and radial temperature and concentration Pexidr. Cerny and Pasek (1968) measured have been measured duri ng catalyst deactivation. +fles axial temperature and concentration profiles for the benzene hydrogenation on a Ni catalyst during Weng. Eigenberger and Butt (1975) studied the deactivation of a non-isothercli;ctivation by CS2. feed. non-adiabatic fixed-bed reactor for the same reaction during thiophene addition to the for this reason no radial temperature The employed reactor had a small inner diameter (0.8 cm): Price and Butt (1977) studied the factors which play a decisive role for an a gradient existed. Richardson (1971) and Wheeler and priori simulation of the reactor behavior during deactivation. parameRobe11 (1969) showed that for isothermal reactors with constant overall deactivation rate ters the activity profiles could be described by the adsorption theory of Bohart and Adam (1920). In the present work we have measured axial and radial temperature and concentration profiles in a An efficient modelling of the observed reactor during deactivation by feed poisoning. fixed-bed mean was achieved by means of a simple pseudo-homogeneous one-dimensional model using a profiles The basic concept is to estimate kinetic and heat transfer parameters cross-sectional temperature. and to assume that the only influence of thiophene adsorption is to from profiles. stationary Thus the deactivation rate parameter can be deterreduce the active surface area of the catalyst. mined by evaluating the profiles taken at a series of times after initiation of deactivation.
EXPERIMENTAL Fig. 1). It The fixed-bed reactor pilot plant has been described by Baiker and Bergougnan (1985a. A jacketed reactor tube of 2.2 m wa* operated as a continuous reactor with recycle of hydrogen. A diagram of the reactor showing the positions of length and 0.052 m inner diameter was employed. (1986. Fig. 1). The catalyst bed and measuring devices has been given by Baiker and Epple the catalyst bed of 0.7 m length was embedded in inert glass packings on both sides. Axial temperature separated by seven positions in the catalyst bed and concentration profiles were measured at 1 and 2 0.1 m ( position 3: z = 0, catalyst bed entrance: position 9: z = 0.6 m: positions The radial temperature and were located in the inert glass bead packing at the reactor entrance). profiles were measured at five radial positions at z = 0.25 m and z = 0.55 m. concentration infrared wall temperature was controlled by circulating oil through the jacket. An The reactor
779
A.
780
BAIKER
D-5
etai.
anaiyzer (URAS) was utilized for the automatic analysis of the toluene concentrations. The plant acquisition were controlled by a PDP ll/lO process A commercial operation and data computer. Ni catalyst (ICI 42.1 ) was used. Properties of the cylindrical (3.4 x 3.4 mm) silica supported pellets were given elsewhere (Baiker and Epple, 11986). Before use the catalyst was catalyst reduced according to the procedure given by the manufacturer. Toluene (99.5 X). nitrogen (99.99%). without further purification. and hydrogen (99.99%) were used Recording of the axial and radial profiles was started after steady-state conditions were reached in the reactor, typically after less than 20 minutes. Then poisoning by addition of thiophene to the reactor feed was initiated. Results of four deactivation experiments performed at 2 bar are presented in this paper. They were characterized by the following conditions. Run (A):Fte;pe;;; ture of oil in reactor jacket. TB = 381 K: total flow rate (hydrogen and toluene). moles/h: Reynolds number, Re = 99: 2.4 vol-%; toluene concentration. thiophene feed, 12 ml/h; total. &a: TB= 393 K; F = 850 moles/h: Re = 104: toluene concentration, 1.8 ~01%; thiophene feed. 25 ml/h. TRE, = o:392 K; toluene concentration, 1.5 ~01%: $nmtFi: F = 690 moles/h, . Re = 78: TB = 394 K: F = 640 moles/h: thiophene feed. Re = 100: toluene concentration. 2.8 vol%: feed. 26 ml/h. thiophene The thiophene adsorption capacity of the catalyst at full deactivation was measured by means of the method described by Granatelli (1959). in which the amount of sulfur is determined. A mean sulfur content of 2.6 mg sulfur per g of catalyst was found. This corresponds to a thiophene adsorption capacity of 0.08 moles per g of catalyst.
MEASURED
AXIAL
PROFILES
Figure 1 depicts the time profiles of temperatures and toluene concentrations and Epple. axial positions 3 - 9 (Fig. 1, Baiker 1986) during deactivation four-point interpolation procedure used to derive continuous axial and radial measured data has been described by Baiker and Bergougnan (1985a).
150
100
Time
Fig.
1
Measured time profiles at positions 3 - 9.
Run
of axial (A).
measured experiment profiles
200
(min)
temperatures
and
toluene
concentrations
at (A). from
the The the
781
Catalyst deactivation of a fixed-bed reactor
D-S
Note that in the initial period Figure 2 shows the axial profiles of experiments (A) and (8). deactivation (t = 5 min) the hot spot temperature was significantly higher than the one of the hot spot shifted towards the With progressing deactivation. steady-state profiles (t = 0). decreased. and the hot spot temperature the profiles were broadened, of the catalyst bed, an inactive zone at the entrance corresponding axial toluene concentration profiles indicated conversion decreased to about which is formed relatively late, i. e. after the catalyst bed, of its initial value.
Catalyst
bed 2
packing
Inert
Temperature
+
bed
Catalyst
I
1
I
Temperature -
1
1
profiles
I
Fig.
2
I 0.7
0.0 axial
I
I
coordinate
z (ml
Interpolated axial temperature and toluene concentration profiles before of runs (A) and (B). Curve 1 corresponds to steady-state deactivation. Curves 2 - 9 represent reactor profiles during deactivation measured in intervals of five minutes after initiation ranges: temperature. of thiophene addition to the feed. Variable 0 - 250 OC: toluene concentration, 0 - 3 ~01%.
of the end The of 20%
A.
782 MEASURED
RADIAL
BATKER
et al.
D-5
PROFILES
The radial temperature profiles measured during deactivation experiments (Fig. 3) were similar in shape to those found during steady-state operation. for which a Biot number h R / k = 5.25 was As a consequence of the parabolic shape found in all experiments.wmean c%oss-sectiocalculated. temperatures could be defined and applied for the reactor model. nal The procedure used for the estimation of the mean cross-sectional temperature has been described by Baiker and Epple (1986). Measured radial toluene concentration profiles were f1at.i.e. the mean cross section concentrations corresponded approximately to the concentrations in the center of the reactor tube.
100
C
100
MINUTES . cl . V m .
IC
1
0 4 8 12 10 28
MINUTES V 0
6 12
L
= 0.55
M
1 (r/R)
100 ID
(r/R)
v 0 l
v A = *
I
I
MINUTES 0 10 12 14 16 18 26
z = 0.25
D
MINUTES 7
10
0
8
m
0
M
1 (r/R)
Fig.
3
Typical radial z - 0.25 m and (Runs C and D;
1 (r/R)
temperature profiles measured at axial positions z = 0.55 m during deactivation experiments. TB = temperature of oil in reactor jacket)
D-5
Catalyst REACTOR
deactivation
of a fixed-bed
reactor
783
MODEL
The reactor model is based on the assumption that the kinetic parameters of toluene hydrogenation thus the kinetic equations for toluene hydrogenation are not affected by the deactivation process: and for the catalyst deactivation are separable. Consequently the heat transfer parameter and the rate constant could be estimated from the initial steady-state of the reactor. In hydrogenation order the deactivation behavior. the model was extended by introducing a linear to describe equation differential equation for the poisoning of the active surface area, and a mass balance for thiophene. The one-dimensional vation is given Heat
pseudo-homogeneous the following set
by
reactor model equations.
of
E
u
Pf
2 lJ (Tb
aT
Cpf
-
T)
+ at
(l-s)Ps
ac
CPS az
R
Boundary
ac
(1-E)
a2
c
the
behavior
without
(l-s)~s
c
Ps
AH r -rr(T.
PT)
deacti-
(1)
=Ps
conditions:
Initial
conditions:
The kinetic investigation
rate (Baiker
r(T.PT)
T(z
= 0.
t)
= Tin(t)
C(z
= 0.
t.)
=
T(z.
t
= 0)
= Tin(t
C(z.
t
= 0)
= 0
= 0)
the toluene 1985b).
(P,/FT)
of catalyst
expIK8
the reactor was defined
(l/T
-
=
pthio
The rate parameter mean temperature T between the estimatzd The second relationship
thio’dt
-
-
been
discussed
in
adsorbed
at
adsorbed
at
total
the
deactivation,
time-dependent
e(t)
k(T,)
k(T) = k according parameters required
u dC thio/da
+
[(l-
exp[-Es/R
(l/T
-
~1 Ps
MT
active
frac-
(4)
Kul’kova (1962) This suggests
have that
found a linear dependence of the time-dependence of g(t)
l/Tm)l.
exp[-E /(RT)], of to theaprocedure of k(T ) and E . to !escribeadeactivation
c)/
earlier
t deactivation
and e(t).
an
(3)
during equation
Andreeva Richardson (1971) and Lyubarskii. the deactivation rate on the active fraction is governed by an equation of the form
/ dt
has
1 thiophene
-d e(t)
hydrogenation
l/Kg))
behavior by the
thiophene =
Cin(t>
=
description of the
e(t)
(2)
r(TvPT)
equation used for and Bergougnan.
1 f
the O(t)
Ps
u--
-
at
dC
describe
bal ante:
m=
For tion
to
bal ante: aT
Mass
used
dimensions Himmelb’lau
d g(t)/dt.
is
bar-’ (1970). the
s-‘,
to
thiophene
has been reduce mass
transformed to the correlation
balance,
a
A.
BAIKER
used
(t
784 initial to the
The following adding thiophene
conditions feed):
were
T(z.
t
=
0)
=
T(z)
(temperature
C(z,
t
=
0)
=
C(z)
(toluene
t
‘thio(” O(z.
t
=
=
0)
Furthermore
0) =
the
=
0
boundary
conditions
=
0.
t)
=
Tin
(temperature
C(z
=
0.
t.)
=
Gin(t)
(toluene
Numerical
=
o.t)
=
(t)
Cthio
is
the
state in
time
before steady-state
when
deactivation
is
started
by
deactivation) before
deactivation
activity).
T(z
C thio(z
steady
0
D-S
concentration)
(initial
following
=
concentration
(thiophene
1
in
et al.
(thiophene
were at
applied:
entrance
concentration
to in
concentration
catalyst
bed)
feed) in
feed)
Solution
orthogonal collocation (Villadsen and Michelsen. The method of 1978) was used to reduce the partial differential equations to a set of ordinary differential equations. In the present work 8 The resulting system consists of 24 differential collocation points were equations, 8 used. algebraic equations for the deactivation, and the deactivation equation at bed entrance (z = 0). This system of equations was simultaneously integrated using Gear’s variable i ntegrati on step These integration routines were used together algorithm, as provided in the software package IMSL. with the nonlinear multiresponse regression programm RKPES (Klaus and Rippin. 1979). This program is based on the “maximum 1 i kel i hood” method (Bard, 1974). RKPES was used with the option of an The experimentally measured data were interpolated with unknown non-diagonal covariance matrix. since the positions of respect to time and space previous to use in the regression program, the measurements did not correspond to the positions of the collocation points, simultaneous and a recording of all measurement points was not possible. Estimated
Parameters
from
Steady-State
Profiles
The kinetic experiments 8.8 -I_ 1.1 to depend
parameters and the overall heat transfer coefficient were estimated from steady state and reported elsewhere (Baiker and Epple. 1986). Mean parameter values found were: k,. mole /kgs: 4.25 + 0.03 K. The heat transfer parameter U was found 7.3 + 1.8 K: k3, on the Reynol kg’s number (cf. Fig. 4 in BTiker and Epple. 1986).
Estimated
Deactivation
Parameters
Deactivation parameters k(T ) and procedure described above. Thg data 30. and 40 minutes CfteT initiation ; E /R. 530 k(T >, 1.05 r. 0.08 s whi!!!h is due to the small $bsolute kJ/mole indicates a weak temperature
were estimated from eight deactivation experiments E by the s& used consisted of the profiles determined at times 10. 20, of the deactivation process. Mean parameter values found were: + 145 K. Note the relatively 1 arge standard deviation of E /R, value of the activation energy. The activation energy of a4. 7 dependence of the deactivation.
DISCUSSION Calculated
and
Experimental
Profiles
Experimental and calculated profiles of runs (A) and (B) are compared in Fig. 4The model provides a good representation of the measured temperature and concentration profiles. The deviation experimental and simulated axial concentration profiles in run (A) at the reactor between outlet can be ascribed to a systematic error in the measurement of very low toluene concentrations. Deactivation
kinetics
The quality of the description of the reactor behavior showed a different sensitivity on the two It was very sensitive to the pre-exponential estimated deactivation parameters. factor k(T ). which in turn was found to be strongly correlated with the thiophene adsorption capacity. In c%the parameter E had only a minor influence on the quality of the fit. which indicates a trast. TRis activation energy of the deactivation kinetics, i.e. very shallow optimum. adsorption the rate of thiophene on Ni. Our value derived from the measured profiles is quite low (4.7 kJ/mole). agrees well with direct determinations in the literature. Eigenberger and Butt (1975) found Weng. while Zrncevic and Gomzi (1983) gave a value of 5.3 kJ/mole. a value of 4.5 kJ/mole. Considering the fact that the thiophene adsorption rate depends only weakly on temperature, an isothermal treatment of the deactivation reaction may lead to satisfactory results. No tion will
definite rate only
conclusions have and the deactivation lead to significant
been
reached concerning rate can be considered errors in the description
the to of
question whether the toluene be separable. However this the profiles when the catalyst
hydrogenaassumption activity
Catalyst
D-5
deactivation
of a fixed-bed
reactor
785
when the conversion in the reactor has dropped to values less than 20 % of the is low. i.e. initial conversion. In contrast to Weng. Eigenberger and Butt (1975) who used a semi-empirical two-s i te mechanism for the deactivation kinetics, we have employed a simple linear dependence between the toluene hydrogenation activity and the adsorbed amount of thiophene. This leads to a considerable reduction of computation time. Consequences of this simplifying assumption are noticeable only for high degrees of deactivation, which are generally of less importance in practical It appears that under the conditions given,the one-dimensional applications. pseudo-homogeneous reactor model can be considered as accurate enough for most practical purposes.
Experimental Temoerature
A
>alculated
orofiles
A
Temperature
profiles
?I
Experimental
Concentration
6
Calculated
pro
6
Concentration
profiles
I
;1
0
L Temperature
r
I
profiles
I
I
axial 4
’
I
’
I
,
coordinate
z
(m
profiles
I
0.7
0.c )
Fig.
perature
I
I
I
0.7
)
Experimental and calculated axial temperature and toluene concentration files of runs (A) and (B). Symbols in calculated curves represent T and values calculated at collocation points. Variable ranges: temperature, 100 - 250 OC: toluene concentration. 0 Curves 1. 2. 3. and 4 correspond to times of 10. 20. 30. and 40 minutes, after initiation of thiophene addition to the feed. respectivley.
pro-
C 3
~01%.
786
A.
D-5
BAIKER etal.
CONCLUSIONS The behavior of a fixed-bed reactor during deactivation has been investigated. Measured axi al prof i les were described by a temperature and concentration pseudo-homogeneous one-dimensional model. Descripton with a one-dimensional model was appropriate because the measured radial temperature profiles exhibited a parabolic shape during the entire course of deactivation. The use of a temperature is justified if the radial position corresponding to mean cross-sectional the mean This condition is fulfilled for radial profiles of the form cross-section temperRt.ure is constant. T(r) = T(0) + a(r/R) . A first order rate expression was found to be appropriate to describe the deactivation caused by thiophene adsorption on the supported Ni catalyst. The thiophene adsorption rate was found to be only weakly dependent on temperature, indicating that in several situations of practical i mportance the use of an isothermal deactivation kinetics may be sufficiently accurate. This simplification is very well justified for moderately deactivated catalyst beds. At high degrees of deactivation a more complex two-site model may be more appropriate.
LIST C r
Lthio cPs ZPf
Gp OH
Re
F
t
T U U z
E
9
3S
OF
SYMBOLS
-3 toluene concentration. mol m -3 thiophene concentration. mol m specific heat of catalyst, J kg-‘K-l specific heat of fluid phase, J kg-‘K-l catalyst particle diaye:fr, m mass flow rate, kg s m reaction enthalpy. J mol-’ heat transfer coefficient throygh,;etctor radial heat conductivity, J m s kinetic parameter, mol/kg kinetic parameter, K kinetic parameter, K -1 thiophene adsorption capacity. mol kg toluene partial pressure. bar mean toluene partial pressure. bar radius of reactor tube. m Reynolds number (G dp/u) radial coordinate, m _, -1 reaction rate, mol kg s time, s temperature, K J ,-zs-1 overall heat transfer coefficipnt. superficial gas velocity, m s axial coordinate, m bed porosity active fraction of catalyst syrface density of fluid phase. kg m-3 density of solid phase, kg m
wall,
J
m-2s-1
K--l
K-l
REFERENCES and M. Bergougnan (1985). Can. J. Chem. Eng.. 63. 138-145. Bai ker, A., and M. Bergougnan (1985). Can. J. Chem. Eng.. 63. 146-154. Bai ker. A., and D. Epple (1986). Appl. Catal., In press. Bai ker, A., Nonlinear Parameter Estimation. Academic Press, New York. Bard, Y. (1 974). and Adam, E.Q. (1920). J. Am. Chem. Sot., 42. 523-544. Bohart, G. S (~~8~~~5~atalyst Deactivation. Elsevier. Amsterdam. Butt, J.B. ) Anal. Chem., 31. 434-436. Granatelli, Himmelblau. D: M. (1950). Process Analysis by Statistical Methods. Wiley. NE!w York. (1984). Deactivation of Catalysts. Academic Press, New York. Hughes. R. and D. W. T. Rippin(1979). Computers & Chemical Engineering 3. Klaus. R.. 105-115. L. B. Andreeva, and N. V. Ku1 ‘kova (1962). Kin. Kat.: 3. Lyubarski i, 102-110. G. D.. Pexidr, V. . Pasek (1968). Proc. 4th European Symp. on Chs km. React .ion Eng.. I. Cerny. and I. Br,ussels. 239-251. Price, T.-H.. and-J. B. Butt (1977). Chem. Eng. Sci., 32. 393-412. Richardson. J.T. (1971). 3. Catal.. 1, 130-138. Michelsen (1978). Solution of Differential Equation Models Villadsen. J. V.. and M. L. Polynomial Approximation. Prentice Hall, New Jersey. G. Eigenberger. and J.B. Butt (1975). Chem. Eng. Sci.. 30. 1341-1351. Weng. H. S., Wheeler. A.. and A.J. Robe11 (1969). J. Catal.. 13. 299-305. Zrncevic. S.. and Z. Gomzi (1983). Chem. Eng. Sci.. 38. 1351-1354.