Академический Документы
Профессиональный Документы
Культура Документы
Abstract
Controlled oxidation reactions of hydrocarbons are practiced today in "xed-bed, #uid bed or transport bed reactors with oxygen or
air as the oxidant. In "xed-bed reactors, due to the exothermicity of the oxidation reactions, heat removal and temperature control are
critical in achieving safe and optimum reaction conditions that maximize conversion and selectivity. Industrial reactors often control
hot spot formation by using an appropriate diluent (ballast) gas. Steam, nitrogen, methane and carbon dioxide are the most
commonly used ballast gases. The scope of this paper is to elucidate the bene"ts of ballast gas use as a tool to optimize heat transfer
and reduce the hot spot e!ect in controlled oxidations with "xed-bed reactors. Ballast gas does not participate in the reaction, but it is
used to control heat removal and #ammability. A homogeneous one-dimensional reactor model was used to study and compare two
oxidation processes: (1) ethylene oxidation to ethylene oxide and (2) o-xylene oxidation to phthalic anhydride. For both processes hot
spot temperature and conversion decreased as methane replaced nitrogen in the process ballast gas. Selectivity increased in the
ethylene oxide case but decreased in the phthalic anhydride case, indicating that ballast gas e!ect on selectivity depends on the
reaction mechanism. Thus, there is an optimum ballast gas composition that will optimize hot spot temperature and reactor yield.
1999 Elsevier Science Ltd. All rights reserved.
Keywords: Ballast gas; Fixed bed; Vapor-phase oxidation; Ethylene oxide; Phthalic anhydride
0009-2509/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 8 ) 0 0 4 9 2 - 8
3684 V. Papavassiliou, M.L. Wagner/Chemical Engineering Science 54 (1999) 3683}3689
Etzkorn and Harkreader, 1988; Etzkorn and Harkreader, heat removal. Oxygen or oxygen-enriched air can be used
1990) describes the use of ballast gas in various hydro- as the oxidant. When enriched air was used, nitrogen was
carbon controlled oxidation processes. We discuss next partially replaced by methane and CO . When oxygen
a few examples of ballast gas use. Our intent is not to be was used, nitrogen was completely replaced by methane
exhaustive on the matter, but rather to present the diver- and CO . Both ballast gas and unreacted butane can be
sity of catalytic oxidation processes that advocate the use recycled. An oxygen-based process with recycle was also
of ballast gases. described (Marshall, 1975; Palmer and Holzhauer, 1982)
Ethylene oxide production in "xed-bed reactors is cur- for the production of maleic anhydride by butane selec-
rently the largest oxygen-based hydrocarbon oxidation tive oxidation. The oxygen-based process was compared
process. Ethylene oxide production was converted from (Palmer and Holzhauer, 1982) to an air-based process
air-based to oxygen-based process in the early 1960s. with a recycle and to the conventional air-based process
Methane (Kingsley and Cleland, 1964) was selected as the using the same catalyst. When pure oxygen was used the
most appropriate ballast gas on the base of its inertness ballast gas was comprised of carbon oxides (CO and
under the reaction conditions, good thermal conductivity CO ), whereas, when air was used, the ballast gas was
and heat capacity properties, and its positive e!ect on comprised of nitrogen and carbon oxides. In order to
#ammability. It has been reported (Kingsley and Cleland, prevent build-up of inerts (CO , CO and N ) in the
1964) that a maximum ethylene oxide yield achieved as recycle a purge stream was withdrawn from the recycle
methane replaced nitrogen in the process ballast gas. The stream at a rate equal to the rate of the build-up of inerts.
optimum yield depended on the ballast gas composition The air-based process with recycle did not o!er signi"-
and the ethylene concentration in the reactor feed. The cant advantages. However, Palmer and Holzhauer
reason for the yield increase was the improved heat (1982), report that the per gram production of catalyst
transfer properties of the methane ballast gas, which was increased by 50% for an oxygen-based process using
reduced the hot spot e!ect and increase selectivity. Vinyl a ballast gas (CO and CO ) compared to the air-based
acetate manufacture is also oxygen-based and alternative process.
ballast gases have been proposed (Severs, 1974) for this
process as well. The current technology uses an excess of 1.2. Heat and mass transfer modeling
ethylene in the reactor feed instead of a non-reacting
alternate ballast gas. The unreacted ethylene is recycled To our knowledge, limited attempts have been re-
and the excess ethylene concentration in the reactor feed ported (Runhong and Yuen, 1993) to elucidate the e!ect
does not adversely a!ect the reaction kinetics for this of alternate ballast gases with reactor modeling tech-
particular reaction. The same solution is also preferred niques. This is surprising given the industrial signi"cance
(Yen et al., 1982) in the oxygen-based chlorination of of ballast gas applications. Runhong and Yuen (1993),
ethylene to vinyl chloride. Oxygen-based vinyl chloride report simulations with one- and two-dimensional
production is an improvement over the air-based pro- homogeneous models of an ethylene oxide reactor when
duction in part, due to better heat transfer properties of nitrogen or methane are used as ballast gases. In reactor
the reaction mixture. simulations, methane performed better than nitrogen by
In most air-based oxidation processes, nitrogen is the lowering the hot spot temperature maximum without
ballast gas. Acrylic and methacrylic acid production are a selectivity penalty. Runhong and Yuen (1993), used
air-based processes that utilize steam as a component of either nitrogen or methane as a ballast gas with di!erent
the ballast gas (Farhad, 1987). It has been reported in operating conditions in each case. Thus, the e!ect on
recent patents (Etzkorn and Harkreader, 1988, 1990) that reactor performance of gradually substituting nitrogen
the use of an anhydrous ballast gas improves production. with methane under identical process conditions was not
The heat capacity of the ballast gas was correlated obvious.
(Etzkorn and Harkreader, 1988, 1990) with improved In recent years increasingly sophisticated models have
reactor performance. Hydrocarbons with higher heat ca- been developed (Finlayson and Rosendal, 1995; Froment
pacity than steam were found to improve performance. and Papageorgiou, 1995; Bey and Eigenberger, 1997a) for
Additional bene"ts included reduction of the maximum heat and mass transfer modeling of "xed-bed reactors.
hot spot temperature and increased catalyst life. Selection of the appropriate reaction kinetic and heat
In order to increase the hydrocarbon concentration in transfer parameters, to be used with a theoretical model,
the reactor feed for maleic anhydride production without is critical. It still remains di$cult to estimate heat transfer
the adverse hot spot e!ects, Ramachaadran et al. (1991) coe$cients (Froment and De Wachs, 1972; Li and Fin-
proposed to use a mixture of methane, nitrogen and CO layson, 1977; Bauer and Schlunder, 1978a,b; Dixon and
as a ballast gas in the feed together with oxygen and the Cresswell, 1979; Cresswell et al., 1982; Wakao and
hydrocarbon. The particular mixture of methane}nitro- Kaguei, 1982; Dixon, 1985; Borkink and Westerterp, 1992;
gen}CO acts as a #ame suppressor allowing higher Bey and Eigenberger, 1997b) in a universally accepted
butane concentration to be used in the feed and facilitates fashion. The problem becomes even more complicated
V. Papavassiliou, M.L. Wagner/Chemical Engineering Science 54 (1999) 3683}3689 3685
increased heat capacity and thermal conductivity on (1977), because that correlation was also used in the
a and j was investigated using empirical or semiem- derivation of kinetic parameters (Runhong and Yuen,
U CP
pirical correlations (Froment and De Wachs, 1972; Bey 1993). For the phthalic anhydride simulation we used
and Eigenberger, 1997; Westerterp, 1997; Wakao and correlations by Wakao and Kaguei (1982), because it was
Kaguei, 1982; Bauer and Schlunder, 1978a, b; Cresswell found that when they were used in simulations the results
et al., 1982; Li and Finlason, 1977). It was found that agree with results reported by Calderbank et al. (1977).
di!erent correlations predict di!erent values for j and The correlations by Wakao and Kaguei can also predict
CP
a , when ballast gas composition changes. Correlations j and a when the composition of the ballast gas
U CP U
based on the Re number (Froment and De Wachs, 1972; changes.
Froment and Bischo!, 1990), though more accurate for
the gas mixture they were derived for, are not suitable to 3.1. Ethylene oxide reactor simulations
predict j and a when the gas mixture composition
CP U
changes. Table 2 summarizes results of several correla- With the aid of the model we tested how ethylene oxide
tions (Bey and Eigenberger, 1997; Westerterp, 1997; yield can be a!ected by the ballast gas composition.
Wakao and Kaguei, 1982; Bauer and Schlunder, 1978a, b; Fig. 1 illustrates the model results for ethylene oxide
Cresswell et al., 1982; Li and Finlason, 1977), incorporat- selectivity, ethylene conversion and hot spot temperature
ing the Pr number. Heat transfer parameters j and as a function of increasing methane concentration. The
CP
a were predicted for a mixture containing initially 18% reactor feed consisted of 15% ethylene, 7% oxygen,
U
molar oxygen, 1% xylene and balance nitrogen. The 10.55% carbon dioxide with nitrogen or nitrogen}meth-
ballast gas molar composition was varied by replacing ane mixtures make up the rest. The reactor feed temper-
nitrogen with methane in 10% increments and monitor- ature was at 5133K and the coolant temperature was
ing the change in the j and a values. The parameters 5263K. It becomes evident (Fig. 1 left) that the selectivity
CP U
for the phthalic anhydride reactor reported in Table 1 (per pass) increases as methane substitutes for nitrogen
were used in the calculation of j and a . The last line in until it exponentially reaches a limiting value. The simu-
CP U
Table 2 represents the average % increase of j and lations predict that ethylene conversion per pass de-
CP
a for a 1% increase of the methane concentration. creases (Fig. 1 left). In the industrial process most of the
U
Although there is disagreement among di!erent correla- unreacted ethylene is recycled, thus, a selectivity increase
tions (Table 2) they all predict a signi"cant improvement will lead to an overall yield increase for the process.
in j and a equal to about 0.14}1.28 and 0.9}1.27%, Fig. 1 (right) also depicts the hot spot temperature evolu-
CP U
respectively, for every 1% methane concentration in- tion as methane replaces nitrogen. We observe that the
crease. Heat capacity and thermal conductivity of the maximum temperature decreases as methane concentra-
gas mixture increase linearly as nitrogen concentration tion in the reaction mixture increases. Reduction of the
decreases and as a result j and a increase. The Bey and hot spot temperature results in a selectivity increase
CP U
Eigenberger (1997) correlation, which has a week (Pr) because higher temperatures favor the undesirable reac-
dependence to the Pr number, predicted the smallest tion to carbon dioxide. These results agree qualitatively
a increase. More work is needed to incorporate the with experimental results presented elsewhere (Kingsley
U
ballast gas e!ect in j and a correlations. For the and Cleland, 1964) for the ethylene oxide process yield as
CP U
ethylene oxide simulations reported in this work we used methane replaces nitrogen in the ballast gas. The experi-
correlations for j and a reported by Li and Finlayson mentally observed yield maximum (Kingsley and Cleland,
CP U
Table 2
Prediction of the wall heat transfer coe$cient (a ) and the e!ective radial conductivity (j ) from various correlations as methane substitutes nitrogen
U CP
in a mixture containing initially 18% oxygen, 1% o-xylene and balance nitrogen
a j a j a j a j a j a j
U CP U CP U CP U CP U CP U CP
0% CH 275 0.74 248 205 0.53 0.97 408 1.10 120 1.70
10% CH 300 0.83 251 229 0.58 1.10 439 1.20 134 1.90
20% CH 324 0.92 254 254 0.64 1.20 471 1.30 148 2.10
30% CH 349 1.00 258 280 0.70 1.30 504 1.40 163 2.30
40% CH 374 1.12 262 309 0.77 1.42 539 1.54 179 2.53
% increase 0.90 1.28 0.14 1.27 1.13 1.16 0.8 1.00 1.23 1.22
per 1% CH
V. Papavassiliou, M.L. Wagner/Chemical Engineering Science 54 (1999) 3683}3689 3687
Fig. 1. Model results for ethylene oxide selectivity, ethylene conversion (left) and hot spot temperature (right) as a function of increasing methane
concentration.
1964) is a function of several process parameters (per pass transfer control in an industrial controlled oxidation
yield, separation system, recycle, purge rate, reactants process.
purity), therefore, a direct quantitative comparison with Fig. 2 presents calculations for o-xylene partial oxida-
experimental results is not possible without knowing of tion to PA as a function of the methane concentration in
all the process details. Also kinetic parameters used in the ballast gas. A feed molar concentration 1.5% in
this work (Runhong and Yuen, 1993) are not representa- o-xylene, 18% in O , and balance N and CH mixtures
tive of the catalyst used in the experimental studies was assumed. A total gas feed rate of 1.318 kg/m s was
(Kingsley and Cleland, 1964). The results (Fig. 1), how- assumed. The reactor wall temperature and the feed
ever, demonstrate a positive e!ect on selectivity and hot temperature were 6933K. We observe from Fig. 2 (left)
spot temperature for the methane ballast gas. that xylene conversion and phthalic anhydride selectivity
decrease as methane concentration in the ballast gas
3.2. Phthalic anhydride reactor simulations increases. The hot spot temperature (Fig. 2 right) also
decreases as methane concentration in the ballast gas
Production of phthalic anhydride (PA) via controlled increases. The conversion and the hot spot temperature
oxidation of o-xylene with air is a well studied oxidation trends are similar to those observed in the ethylene case.
process. In phthalic anhydride production process The product selectivity trend (Fig. 2 left), however, is
proper care must be taken to ensure that the hydrocar- opposite to the one observed in the ethylene oxidation
bon-air feed mixture is outside the #ammability envelope case. This di!erence is explained if one considers the
or an elaborate safety system must be employed to pre- more complicated reaction mechanism that we used for
vent and/or relieve a de#agration. Operating within the the xylene oxidation (reaction (III)). In this case not only
#ammability envelope is possible (Lurgi, 1994) by care- the undesirable reaction rates, that form carbon oxides,
fully designing the reactor feed system, by using proper are depressed by the lower temperatures encountered
rupture disks to relieve pressure and by using redundant along the reactor but the reaction rates of the intermedi-
interlocks. The lower #ammability limit of o-xylene in air ate species, are depressed as well.
is 1 mol%. The di!erence in the temperature pro"les along the
The model was used to predict behavior of "xed-bed reactor is depicted in Fig. 3 for nitrogen ballast gas (left)
reactors in the air-based controlled oxidation of o-xylene and for nitrogen with 10% methane ballast gas (right).
to phthalic anhydride, in which the properties of the feed We observe that the temperature maximum is lower
gas mixture were altered by introduction of a methane when the ballast gas contains methane. The temperature
ballast gas. We employed this example for the purpose of maximum that is observed (9003K) when nitrogen only is
illustrating the ballast gas e!ect on heat transfer, how- used, would result in fast catalyst deactivation during
ever, several other factors (e.g. #ammability, catalyst actual reactor operation (Calderbank et al., 1977).
performance, process economics) need to be evaluated Catalyst dilution was proposed as a solution for hot
before one can actually utilize a ballast gas for heat spot temperature control, and was proved e!ective in
3688 V. Papavassiliou, M.L. Wagner/Chemical Engineering Science 54 (1999) 3683}3689
Fig. 2. Model results for phthalic anhydride selectivity, o-xylene conversion (left) and hot spot temperature (right) as a function of increasing methane
concentration.
Fig. 3. Temperature pro"les along the reactor length for phthalic anhydride when nitrogen is used as a ballast gas (left) and when a nitrogen with 10%
methane mixture is used as ballast gas (right).
theoretical simulations (Calderbank et al., 1977). The hot spot temperature control in "xed-bed reactor pro-
hot spot temperature maximum is reduced from 9003K cesses. Di!erent reaction mechanisms may be a!ected
(Fig. 3 left) to 8503K (Fig. 3 right) when a ballast gas di!erently by changes in the ballast gas composition as it
containing 10% methane is used. Although conversion is evident from the comparison of the ethylene oxidation
and selectivity are somewhat lower the catalyst life will be and the o-xylene oxidation cases. In this work the e!ect
preserved. Further increase of the methane concentration of a single ballast gas was examined, although, the ballast
in the ballast gas will further reduce the hot spot temper- gas composition is a critical parameter that requires
ature but also reduce product yield. Thus, for a given set optimization. In an industrial reactor the ballast gas
of process conditions, there is a methane/nitrogen ballast selection will be guided by reactor performance evalu-
gas composition, that will optimize both hot spot tem- ation, process parameters (recycle, purge, #ammability)
perature and phthalic anhydride production. and economic optimization. Therefore, it may not always
be economical to use the best ballast gas available.
Future work should focus on the development of reli-
4. Conclusions able heat transfer correlations that account for composi-
tion changes, and also on the study of the ballast gas
Theoretical results from ethylene oxide and phthalic e!ect with heterogeneous models, in order to provide
anhydride reactor simulations indicate that ballast gas accurate design guidelines for reaction and process
use is an e!ective tool for improving heat removal and engineers.
V. Papavassiliou, M.L. Wagner/Chemical Engineering Science 54 (1999) 3683}3689 3689
Notation Borkink, J.G., & Westerterp, K.R. (1992). Chem. Engng Sci., 47,
2337}2342.
a wall heat transfer coe$cient Calderbank, P.H., Chandrasekharan, K., & Fumagalli, C. (1977). Chem.
U Engng Sci., 32, 1435}1443.
Bi Biot number Cresswell, D.L., Newson, E.J., & Wellauer, T. (1982). Chemical reaction
c heat capacity
N engineering, ACS Symp. Ser. No. 196, (pp. 527}542).
d catalyst particle diameter Dixon, A.G., & Cresswell, D.L. (1979). A.I.Ch.E. Jr., 25, 663}676.
N Dixon, A.G. (1985). A.I.Ch.E. Jr., 31, 826}834.
G mass velocity
H heat of reaction Etzkorn, W.G., & Harkreader, G.G. (1988). European Patent,
293224A1.
¸ reactor length Etzkorn, W.G., & Harkreader, G.G. (1990). European Patent,
M average molecular weight
361372A1.
P total pressure Farhad, N. (1987). Acrylic acids and acrylic esters. SRI report 6C.
Finlayson, B.A., & Rosendal, B. (1995). Comput. Chem. Engng, 19,
r reaction rate
r reactor tube radius 1207}1218.
R Froment, G.F., & De Wachs, (1972). Chem. Engng Sci., 27, 567}576.
Re Reynolds number Froment, G.F., & Bischo!, K.B. (1990). Chemical reactor analysis and
¹ temperature design. New York: Wiley.
X conversion Froment, G.F., & Papageorgiou, J.N. (1995). Chem. Engng Sci., 50,
y feed mole fraction of hydrocarbon 3043}3056.
Harold, M.P., & Lerou, J.J., & Mills, P.L. (1995). Catalytic oxidations,
e catalytic bed porosity
(pp. 291}369).
j e!ective radial conductivity
CP Herskowitz, M., & Hagan, P.S. (1988). A.I.Ch.E. Jr., 34(8).
Herskowitz, M., Hagan, P.S., & Pirkle, C. (1988). SIAM Jr. Appl. Math.,
Subscripts 48, 1083}1101.
Kingsley, H.A., & Cleland, F.A. (1964). ;.S. Patent 3119837.
i species Li, C.H., & Finlayson, B.A. (1977). Chem. Engng Sci., 32, 1055}1066.
Lurgi, A.G. (1994). Lurgi brochure, Phthalic anhydride.
w wall Marshall, E.F. (1975). ;S Patent 3904652.
r reaction averaged Palmer, D., & Holzhauer, J. (1982). ;S patent 4342699.
t tube Perry, R.H., & Green, D. (1984). Perry's chemical engineering hand-
book. (6th ed.). New York: McGraw Hill.
Ramachaadran, R., Shukla, Y., & MacLean, D.L. (1991). ;S Patent
4987239.
References Runhong, Q., & Yuen, T. (1993). Shiyou Huagong (Petrochemical Engin-
eering). (vol. 22, pp. 792}798.
Bauer, R., & Schlunder, E.U. (1978a). Int. Chem. Engng, 18, 181}188. Severs, G.M. (1974). ;S patent 3855280.
Bauer, R., & Schlunder, E.U. (1978b). Int. Chem. Engng, 18, 189}204. Wakao, N., & Kaguei, S. (1982). Heat and mass transfer in packed beds.
Bey, O., & Eigenberger, G. (1997a). Chem. Engng Sci., 52, 1365}1376. London: Gordon and Breach.
Bey, O., & Eigenberger, G. (1997b). Vortrag, ECCE-1, May 4}7. Yen, Y., Hsu, L.C., & Chiou, C. (1982). SRI Report 5C.