Computers in Industry 18 (1992) 213-219
Elsevier
Short Note
Mathematical model of the process for the
oxidation of acetaldehyde to acetic acid
Wang Shuqing, Z h a n g Rufong and Wang Jicheng
Research Institute of Industrial Process Control, Zhejiang Unicersity, Hangzhou 310027, China
Received March 30, 1991
Accepted May 15, 1991
The acetaldehyde oxidation process is one of the important
steps of acetic acid production. The reaction takes place in a
multistage bubble column and the reaction mechanism is very
complicated involving gas-liquid mass transfer, hydro-
dynamics and other factors, and a set of series-parallel reac-
tions. Based on the reaction kinetics, mass transfer and mate-
rial balances, a set of nonlinear equations is obtained that can
be solved by iteration. The ratio of acetaldehyde to oxygen
and the reaction temperature are the main variables which
govern the oxidation reaction in the bubble column. Studying
the influence of these two variables on the oxidation reaction
and comparing the simulation and the experimental data, the
results are satisfactory. Finally, a simple control model and
the optimal operating conditions are obtained, providing a
basis for optimal control of the acetaldehyde oxidation proc-
ess.
Keywords: Oxidation reaction, Gas-liquid reaction, Mod-
elling, Simulation, Optimal control, Control model,
Acetaldehyde oxidation.
1. Introduction
The process for the oxidation of acetaldehyde
to acetic acid is one of the main steps of acetic
acid production. The reaction takes place in a
concurrent multistage bubble column partitioned
by perforated plates. A bubble column is a kind
of multiphase chemical reactor, which is widely
used for gas-liquid and gas-liquid-solid reac-
tions in chemical industry [1]. The gas-liquid
reaction is a very complex process which com-
bines mass transfer and chemical reaction. A lot
0166-3615/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved
213
of work has been done on different types of
bubble reactors with different methods [2], gener-
ally beginning with kinetic studies of the influ-
ence of the transfer process on reaction rate and
selectivity.
There are three kinds of models for a multi-
stage bubble column. The simplest is the continu-
ous stirred tank reactor (CSTR) in series model
[3-5], which treats the influence of backflow with
a pseudo series number or series efficiency. How-
ever, this model is not suitable for reactors with
multichannel feed. Over the last decades, investi-
gators [6,7] have paid more attention to the
back-flow model (BF model). In this model each
stage is completely mixed up, but there exists
back flow between two stages, the amount of
back flow being expressed by a back-flow ratio, a.
The third model is the dispersion-back-flow
model [8]. When the ratio between height and
diameter is large enough at each stage, then the
assumption of complete mixing cannot be used,
and then the dispersion-back-flow model should
be selected. These models, which have consid-
ered unperfect mixing between materials in in-
stallations, can give a better description of the
real flow processes and are suitable to solve the
problem of multi-dimensional and multi-compo-
nent separations and reactions effectively.
The purpose of this research work is to de-
velop a mathematical model of an industrial bub-
ble-column reactor for the production of acetic
214 ShortNote Computers in Industry

acid. Based on the model proposed and industrial
experimental data, the simulation is performed,
and then the optimal control model and operat-
ing conditions of the industrial bubble-column
reactor are given.
2. Mathematical model
2.1. Kinetics
Oxidation of liquid acetaldehyde in the pres-
ence of metal salts such as manganese or cobalt
acetate produces acetic acid according to the
reaction
reaction since 1930. The oxidation of acetalde-
hyde is a typical example of a chain reaction, and
the reaction mechanism can be interpreted satis-
factorily in terms of a free-radical mechanism.
Although there are many research reports [11,12]
on the reaction mechanism and kinetics, these
kinetic equations have not yet been satisfactorily
applied in an industrial process. Gap and Dai [13]
proposed macroscopic kinetic equations based on
Higbie's model of gas-liquid phase reactions. The
kinetic models are much better for use in prac-
tice. The kinetic equations are:
__ ,.
a IfkaDLg p t',o.5
-r°2 k'aP°2C°'5 1 - ¢6 HA o2"n ,

2CH 3caP -I- 0 2 --> 2CH aCOOH. (1) (2)

Extensive studies [9,10] have been done on the ka= 2.402 × 108 e-8"21x lo3/nr. (3)
The process can be simplified as follows:
Wang Shuqing received his BS degree
from the in Chemical Engineering CHaCHO + 10 2
Department of Zhejian University,
China, in 1964. From 1964 to 1985 he
CB(xl)
worked as a lecturer at the Depart-
ment of Chemical Engineering at the k,', ~ CH3CO0 H
same University. He was a visiting
scholar at Loughborough University k'
of Technology, UK, 1985-1986. Since byproduct of gas phase
1986 he is an Associate Professor and
works at the Research Institute of k u
Industrial Process Control, Zhejiang ~ byproduct of liquid phase C~(x 3)
University, He has majored in the
field of automatic control of chemical and biochemical proc- Gianfrauco [14] has given:
esses. He is interested in the modelling and optimization of
industrial processes, and the application of adaptive control,
fuzzy control and expert systems, and the design of DCS and k'-alk ~, (4)
microcomputer control systems,
k " - a 2 k a, (5)
Zhang Rufong was born in Zhejiang,
China, in 1966. He received his BS where a~ and a2 are constants.
and Master degree in industrial proc-
ess control from Zhejiang University, Next the multistage back-flow model is used to
China, in 1986 and 1989 respectively. describe the performance of the multistage bub-
Since 1989 he has worked for Zhe-
jiang Petrolium Chemical Automation ble column (see Fig. 1).
Company, Hangzhou. His research in- The back-flow ratio a'is defined as the ratio of
terest is in the development and ap-
plication of modern control theory. back-flow rate to the net flow rate of liquid, and
He has published five papers. He is a the equation proposed by Sekizawa and Kubota
member of the Hangzhou Automa-
tion Society. [8] is used:

Wang Jicheng is a Professor at the 17.96u~2~ ( 0.01 0.011 )

Chemical Engineering Department,
Zhejiang University, China. He was a a = 1 + uo/rl uL u°'85~/°'2
visiting professor of Lehigh Univer-
sity, USA, 1984-1985. He is inter- (( 2"512 X 10-68°'2 )
ested in process dynamics, optimal
control of chemical processes etc. X 1.48 exp 3.16/~°'5 + do2
Computers in Industry S. Wang et al. / Modelling acetaldehyde oxidation 215

!
UL,XN I FGN'YN bered 1,2 and 3, respectively. The concentration
and reaction rate of the ith component within the
jth stage in the liquid phase are denoted by xji
and rji, respectively.
, f , , I
(I+~j)UL,X j • i 0tj+IUL,Xj+I : i - r j l = - 2ro2j(1 + a I + a2) (16)
F..,Y
j L j I ~ ° rj2 = 2( - rod) (17)
rj3 = 2 ( - r o ~ ) a 2 (18)
- i I 1" FDy - 2a,( - r o d ) ( 1 - EG)Vj, (19)

1 | Interface where FDj is the mole flow of the byproduct in

, mass transfer the gas phase within stage j.
I UL,X
/ F021
DY0 The percentages of the total conversion and
Fig. 1. Back-flow model.
the productivity are given:
Cov = ( Xol -- X N l ) / X o , , (20)
C ~ V -" ( X N 2 -- X02 )//Xol. (21)
So the mass balance equation of each component So, the selectivity is
in the liquid phase around each stage becomes:
SF - C ~ v / C o v (22)
and the raw material consumption of the process
+ aj+ lU'LXj+ l -- R; = O, (7) is obtained as:
which, divided by u[, gives: 44 1
+ + (a Y. . . . . (23)
60 C~v
+%+~xj+~ - R ~ f f i 0 . (8)
This can be written in the form of a matrix: 2.2. Mass-transfer equations
G M ffi A ' x i + B'Xio + R i ffi O, (9) It is assumed that reaction is in the "fast
where reaction regime". We can treat the reaction as a
M T pseudo-second-order irreversible fast reaction,
¢J~ ffi [ G ~ . . . Gff . . . GN, ] , (10) and the gas-phase resistance is considered negli-
gible.
Xi -" [ Xli . . . X j i . . . XNi] r , (11) The oxygen is transferred from the gas phase
to the liquid phase in each stage and the amount
of oxygen (moles) transferred can be calculated
...ON(1 --~G)rNi]T, (12) by the following equation:
Oj - V J u ' L , (13) ( Y ) A'[3PkLaLj(Y-Y*)dz (24)
-(1 +at) -FGd "l-y - HA
I+a I --(1 + 2a 2) a3
At_~ •. ". ". , (14) where FC and y are the flowrate and mole frac-

a2 1 +OtN_ I
aN
--l--a N
I tion of oxygen in the gas phase at arbitrary axial
coordinate, and /3 is the chemical absorption
enhancement factor:
B' = [ 1 0 . . . 0] T. (15) 1
For convenience we assume that there are three 13 = _.~L~/kaDLA X (25)
components, acetaldehyde, acetic acid and the
byproduct in the liquid phase, and they are num- For a multi-stage bubble-column reactor and upon
216 ShortNote Computers in Industry

Table 1 tion process is obtained as equations (9), (26) and

Parameters of the mass-transfer equations (31), and given a set of nonlinear equations to be
Parameter Ref. solved by iteration.
DLA = 8.18X 10-9 e -3131x 103/RT [13]
~rA 1 1"~,,/tn5 =-3.98×103/RT [131
dvs=26dlgD2pLl-o.5(~)-O.12(~D) -0.12 [15] =0 i=1,2,3 j--1,...,N. (32)

[161
The material flow and the composition of the
a = 6ca/dvs 1/3 - 1/2 [16] feed, i.e. P and Jj, within each stage are known,
k L 0.42
= [171 so that there are N(c + 2)variables (i.e.
xji, Foj, y~) left. Equation (32) can be written in
the form of a vector:
integrating eqn. (24), the following equation is
obtained:
g(w) =0, (33)

G~ = A t ~ j k L a L j P / H A where

+ w=[w,...wN] T, (34)
- oj_,[ln y;(1 - y j - , )
1]
1 --Yj-I

1-y;'
(26)
g= ...
wj = [ xj,... jc, F j, yj]
(35)
(36)
where y] is the outlet mole fraction of oxygen
from stage j, and yj is the mole fraction of
gj-[Gj~...G~, G~, G~] T. (37)
oxygen mixed with feed oxygen and the byproduct
The process of calculation can be summarized
of the gas phase.
as follows:
The parameters of the mass transfer equations
are calculated by the equations listed in Table 1.
The material balance of oxygcn for each stage (1) An initial set of values is assumed for x~i, Fo r,
is shown in Fig. 2 and is described by the follow- and y~.
ing equations: (2) Calculate all kinds of parameters in the model
Foj_t = Fo~i + F ~ _ I + FDj_ I, (27) and the reaction rate profile along the col-
umn.
F~j_ I = FOj_ 1 -- Fo2j -- FDj_ I , (28) (3) Calculate the remainder of eqn. (32).
FG#_ lY#_ l = F~j_ ! y]_ l + Fo,jy 0, (29)
Y/-, = (Fo~_Iy~_, - FodYo)/F/3~_,. (30) ! !

The reaction rate of oxygen is equal to the con- ~ FGj'Yj

sumption rate of oxygen in the gas phase, so that
we have:
= F,~j_~y;_ 1 - y j ,F%t
I,I
+ ~(1 - eG)ro2J( 1 + a I + 92) --0. (31)

FO2JsY0 - / Y I i- I
3. Simulation
FGj-I'Yj-1
On the basis of the above discussions the
mathematical model of the acetaldehyde oxida- Fig. 2. Oxygen material balance.
Computers in Industry S. Wang et aL / Modelling acetaldehyde oxidation 217

(4) Calculate the norm of the remainders.

~ T=82 °C
= + + (G,H) . 99 * ~ ~ . ~ ="-.,,...~...._~
jr! i=i
(38) ~ 98 +
The condition tr < E is considered to be crite-
rion of convergence, and E is the desired
accuracy.
(5) Otherwise, extending eqn. (32) by a Taylor o~
series and taking its linear term, the following o96
iteration variables correlation equation is ob-
tained, which can be used for calculating cor-
' " I I '

rect values of the iteration variables: 3. l 3.2 3•3 3.4 3.5 E(Kg/Nm 3)

AW=-
(°,),
~ g (39)
Fig. 3. Simulation results for percent conversion (o, *, + are
experimental data).

Then return to step 2.

The model indicates that the remainder in stage j

only relates to the variables in stages j, j - 1 and
98
j + 1, so 8g/Ow has the form of a tridiagonal
matrix: 97
B i C!
0g
,42 B2 C2 // -"1
.~ I / / / I
l •

(40)
0w •~ " Y / / I
CN- I
~r/- I
AN BN
So simulation calculations of the bubble column
3I! 312 313 314 3'5 E(Kg/ 3)
can be performed•
Based on studies of the physical plant and Fig. 4. Simulation results for selectivity (o, *, + are experi-
practical operation experience, we know that the mental data).
ratio of acetaldehyde to oxygen, E, and the reac-
tion temperature, T, are the main variables gov-
erning the oxidation reaction in the bubble col-
umn. Studing the influence of these two variables
44 I
on the reaction results and comparing simulation Y= "~ COV*SE
and experimental data, the results are satisfactory 0;82
(see figs. 3-5). Through treatment of the simula- >,,
tion results, a simplified control model is ob- T=74 °C
0.80
U
tained: T=82 °C
Y = -6.2531 + 0.04989T + 0.05112E 2 O.78~ . , ~ . .
+ 5.998/E 2 + 12346.84/T 2, (41) T=78 °C

Coy = -4.54321 - 0.066556E 2 + 0.004779ET 0.7b

- 7 2 0 . 3 1 3 8 / T - 25699.16/T 2, (42)
I I I I '
S E - 4.52078 - 0.0001027 T 2 _ 0 •0004498 ET 3.1 3.2 3.3 3.4 3.5 E(Kg/Nm3)
- 7.471/E 2 - 6569.261/T 2. (43) Fig. 5. Simulation results for raw material consumption.
218 Short Note Computers in Industry

The results of simulation show that: (1) When gas-liquid phase interfacial area,
E increases, i.e. acetaldehyde becomes in excess m2/m 3
of oxygen, conversion of acetaldehyde decreases Cov, C )v conversion
gradually, while productivity increases (2) When D column diameter, m
T rises, the reaction is intensified so that conver- DLA diffusion coefficient, m 2 / s
sion increases evidently, and productivity de- do hole diameter of plate, m .
creases. This is due to the fact that the side dvs average bubble diameter, m
reaction is intensified simultaneously. Figure 5 Fo flowrate of byproduct of the gas
shows that there exists an optimal T and an phase, kmol/s
optimal E which minimize the raw material con- flowrate of gas, kmol/s
sumption of the process. Foe flowrate of feed oxygen
The simplified models can be used to realize H liquid height in column, m
optimal control so as to minimize the raw mate- Henry's constant, m 3 bar/kmol
rial consumption of the process. The flexible tol- k reaction rate constant
erance method can easily be applied to the opti- kL physical mass transfer coefficient on
mization problem and gives a reliable result: the liquid side, m / s
Lj jth stage reactor length, m
E* = 3.30 k g / N m 3,
N overall stage number
T * -- 79.7 ° C, P pressure, bar
Y - 0.7743. r reaction rate, k m o l / m 3
S~ productivity
T temperature, K
4. Conclusion volume of bubble column and of
jth stage reacter, m 3
Comparing the simulation with experimental UG superficial velocity of the gas, m / s
data from industry the results show: UL superficial velocity of the liquid,
(a) The mathematic models, equations (9), (26) m/s
and (31) proposed by this paper can be used to flowrate of the liquid, m3/s
describe the cocurrent multistage bubble column X molar concentration of component
of gas-liquid reaction process successfully. in liquid phase, kmol/m 3
(b) Based on the models the influence of the Y raw material consumption in the
ratio of acetaldehyde to oxygen E, and the reac- process
tion temperature T, on conversion Coy, selectiv- mole fraction of oxygen in the gas
ity SE and Y are studied. The results shown in phase
Figs. 3-5 are satisfactory raw material consump- a back-flow ratio
tion. parameter defined in eq. (25)
(c) On-line measurement of the percentage of v surface tension, N / m
conversion Cov and selectivity SE and the raw PL density of liquid phase, k g / m 3
material consumption Y, is unavailable now. rl free area of plate
However, E and T can be measured on-line. So, IJ, L kinematic viscosity of liquid phase,
by using the simulation results of Figs. 3-5, sim- m2/s
ple estimating models of Cov, S E and Y are 6 plate thickness, m
obtained as equations (41), (42) and (43). This is a EG gas hold up
basis to realize optimal control of the acetalde- 0 time, s
hyde oxidation process.

References
Notation
[1] W.D. Deckwer and A. Schumpe, "Bubble columns. The
A! cross section area of bubble col- state of the art and current trends", lnt. Chem. Eng.,
umn, m 2 Vol. 27, 1987, pp. 405-422.
Computers in Industry
[2] Y.T. Shah, G.J. Stiegel and M.M. Sharma, "Backmixing
in gas-liquid reactors", AIChE J., Vol. 24, 1978, pp.
369-400.
[3] Lu Nanchen, "Mathematic model of continuous reaction
process with volatile liquid phase", China Sci., Vol. 1975,
pp. 267-277 (in Chinese).
[4] Shi Lichai, "Evaluating the hydrodynamic model param-
eters of concurrent multistage bubble column partitioned
by perforated plates", Chem. Eng., Vol. 2, 1982, pp.
43-52.
[5] A.A. Shaikh and A. Varma, "Modeling of gas-liquid
continuous-stirred tank reactors (CSTRs)", Am. Chem.
$ci.,Vol. 84, 1984, pp. 95-106.
[6] T. Miyauchi and T. Vermeulen, "Diffusion and backflow
models for two-phase axial dispersion", Ind. Eng. Chem.
Fund., Vol. 2, 1963, pp. 305-311.
[7] W.D. Deckwer and J. Hallensleben, "Exclusion of gas
sparger influence on mass transfer in bubble columns",
Can. Z Chem., Vol. 58, 1980, pp. 190-197.
[8] T. Sekizawa and H. Kubota, "Liquid mixing in multistage
bubble columns", J. Chem. Eng. Jpn., Vol. 7, 1974, pp.
441-446.
[9] C.E.H. Bawn and J.B. Williamson, "The oxidation of
acetaldehyde in solution", Trans. Faraday Soc., Vol. 47,
1951, pp. 721-743.
S. Wang et al. / Modelling acetaldehyde oxidation 219
[I0] C.V. Gurumurthy and V.M.H. Govindarao, "Rate model
and mechanism of liquid-phase oxidation of propionalde-
hyde", Ind. Eng. Chem. Fund., Vol. 13, 1974, pp. 9-17.
[ll] B.H. Carpenter, "Oxidation of acetaldehyde to acetic
anhydride", Ind. Eng. Chem. Proc. Des. Dev., Vol. 4,
1985, pp. 105-111.
[12] B. Venugopal, R. Kumar and N.R. Kuloor, "Oxidation of
acetaldehyde to acetic acid in a sparger reactor", Ind.
Eng. Chem. Proc. Des. Dev., Vol. 6, 1967, pp. 139-146.
[13] F. Gao and J.L. Dai, "A kinetic study of gas-liquid
phase catalytic oxidation of acetaldehyde for ethanoic-
acid systhesis", Chem. Reaction Eng. Technol., Vol. 1,
1985, pp. 11-20 (in Chinese)
[14] G. Oliveri del Castillo, Chem. Ing. Tech., Vol. 55, 1983, p.
889.
[15] K. Akita and F. Yoshida, "Bubble size, interfacial area,
and liquid phase mass transfer coefficient in bubble
columns", Ind. Eng. Chem. Proc. Des. Dev., Vol. 13, 1974,
pp. 84-91.
[16] G.A. Hughmark, "Liquid-liquid spare column drop size,
hold up, and continuous phase mass transfer", Ind. Eng.
Chem. Fund., Voi. 6, 1967, pp. 409-413.
[17] P.H. Calderhank and M.B. Moo-Young, "The continuous
phase heat and mass-transfer properties of dispersions",
Chem. Eng. Sci., Vol. 16 1961, pp. 39-54.