Comparison of Different Kinetic Models for the Chlorohydrin

Process Using Real Data from an Industrial Tubular Reactor

Ivan Petric* and Amel Husanovi
c

Department of Process Engineering, Faculty of Technology, University of Tuzla, Univerzitetska 8 75000, Tuzla, Bosnia and Herzegovina

The main objective of this study was to investigate how different kinetic models for the chlorohydrin process can inuence the simulation results. In
order to accomplish this objective, a simplied reactor model was developed and applied on the kinetic models. The simulation results were compared
with actual process data (three different data sets) from an industrial tubular reactor in order to nd the most appropriate kinetic model as well as to
validate the model. The results of numerical simulations showed good agreement with the measured actual values. The inuences of the reactor
length, the input ow temperature and the entering molar ows of components on the process performance were studied using the validated
mathematical model. The maximum molar ow rate of propylene oxide was achieved with temperature of the entering reaction mixture of 57 8C. The
reactor length, the input ow temperature and the molar ows of components have the strongest inuence on overall selectivity of propylene oxide
related to propylene glycol.

Keywords: kinetic model, chlorohydrin process, tubular reactor, reactor model, numerical simulation

INTRODUCTION Molnar et al.[7] have developed a mathematical model for the

synthesis of propylene oxide in a continuous stirred tank reactor. The

P
ropylene oxide (propene oxide, metyloxirane or 1.2
model was used for determination of multiple stationary states, or to
epoxypropane) is one of the most important starting
determine the accuracy of the model to the analysis of reactor safety.
materials in the chemical industry. Propylene oxide has a
According to our knowledge based on an extensive review of the
worldwide production value of over 8 million tons per year.[1] The
literature, this is the rst study dealing with the comparison of
production of propylene oxide consumes over 10 % of all
different kinetic models for the reaction of synthesis of propylene
propylene produced.[2] The major application of propylene oxide
oxide by chlorohydrin process using real data from an industrial
is in the production of polyether polyols (65 %), which are mainly
tubular reactor. Therefore, the main objective of this study is to
used for the production of polyurethane foams. The second, third
investigate how different kinetic models inuence the numerical
applications of propylene oxide are in the production of propylene
simulations using real data from an industrial reactor. In order to
glycol (30 %), propylene glycol ethers (4 %), respectively.[3]
accomplish this objective, a simplied reactor model will be
Propylene glycols are mainly used in the production of polyesters,
develop and apply on the kinetic models. The simulation results
whereas propylene glycol ethers are primarily used as solvents.
will be compared with actual process data in order to nd the most
Propylene oxide is produced in several different types of
appropriate kinetic model as well as to validate the model. Finally,
processes: the chlorohydrin process, the hydroperoxide process,
the inuence of the reactor length, the input ow temperature
ethene epoxidation process with a silver catalyst, hydrogen peroxide
and the entering molar ows of components on the process
combination process, etc. In practice, the most widely used plants
performance within the reactor will be studied using the validated
are based on the chlorohydrin and the hydroperoxide processes.[4]
mathematical model.
The mathematical model represents useful tool for simulation
and design of process equipment in chemicalprocessing industry,
as well for determination of the conditions for their optimal METHODS
performance. The industrial signicance of such research lies in the
fact that the understanding of the process has been made clearer. The Synthesis of Propylene Oxide
The results of such studies may serve as useful guidelines for The production of propylene oxide in Bosnia and Herzegovina
improvement the plant design and operation in order to improve started in 1976, in the company SodaSo Tuzla. Since then the
the overall process performance without the need for expensive test company name has changed several times over, from Chlor
runs trials in the plant.[5] Alkaly Complex and Polihem to the Poliolchem d.o.o. Tuzla.
The ability of models to simulate the reactor behaviour depends
on the appropriate values of kinetic parameters. For the synthesis
of propylene oxide by chlorohydrin process, several different
kinetic models can be found in available literature.[68] *Author to whom correspondence may be addressed.
There are several studies dealing with mathematical models for Email address: ivan.petric@untz.ba
the synthesis of propylene oxide by chlorohydrin process, but these Can. J. Chem. Eng. 93:7887, 2015
are related to the reactive distillation.[5,810] These studies are 2014 Canadian Society for Chemical Engineering
DOI 10.1002/cjce.22116
about modelling and simulation of process unit which simulta- Published online 12 November 2014 in Wiley Online Library
neously performs chemical reaction and distillation. (wileyonlinelibrary.com).

78 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 93, JANUARY 2015

The golden age of the company was the eighties in the last century.
The highest production was in 1989 with 23 854 tons of produced
propylene oxide.[11]
A small excess of lime is added to the chlorohydrin reactor liquor
to saponify the propylene chlorohydrin to propylene oxide and
neutralize the hydrochloric acid. These reactions are carried out in
the tubular reactorsaponier (Figure 1). From the lime injection
point to the distillation feed pumps, the reactor is 0.20 m in
diameter and 64.9 m in length. A small amount of propylene glycol
is also produced by the sequential reaction of propylene oxide with
water. Hydrochloric acid neutralization takes place rapidly, so the
rst 2.44 m of the reactor is made of titanium and the rest of the
reactor is made of mild steel.
Propylene oxide is formed from propylene chlorohydrin in an
alkaline environment according to the following dehydrochlorina-
tion reaction (also called saponication):
CH 3
CHOH
CH 2 Cl CaOH2 HCl
1 Equations (5) and (6):
! CH 3 CHCH 2 O CaCl2 2H 2 O:
Propylene oxide hydrolysis to propylene glycol occurs as an
undesired consecutive reaction:
CH 3 CHCH 2 O H 2 O ! CH 3
CHOH
CH 2 OH: 2 Equations (5)(12):
Development of Mathematical Model
Assumptions and simplications
The following assumptions and simplications are used in the
derivation of the model equations:
There are no radial gradients of temperature and concentration.
The reactor operates in a steady state conditions.
The heat capacities of calcium hydroxide and calcium chloride
are constant in the temperature range in the process.
The heat (enthalpy) of mixing is negligible.
The change in density of the reaction mixture is negligible.
The effect of pressure drop on the reaction rate can be neglected.
Among the side reactions, the synthesis of propylene glycol is
the only signicant reaction.
Mole balances
Using the notation from Table 1, the stoichiometric Equations (1)
and (2) can be presented in a more convenient way with
A B C ! D E 2F; 5
D F ! G: 6
The mole balances for all components can be represented by

dF A
r A r 1A ; 7
It should be noted that the reaction (1) is obtain by summing the dV
following reactions (3) and (4):
dF B
r B r 1A ; 8
dV
CH 3
CHOH
CH 2 Cl 0:5 CaOH2
3
! CH 3 CHCH 2 O 0:5 CaCl2 H 2 O; dF C
r C r 1A ; 9
dV
0:5 CaOH2 HCl ! 0:5 CaCl2 H 2 O: 4 dF D
r D
r 1A r 2D ; 10
dV

dF E
r E
r 1A ; 11
dV

dF F
r F
2r 1A r 2D ; 12
dV

dF G
r G
r 2D : 13
dV

Kinetics
The rate laws of the rst reaction (Equation (5)) and the second
reaction (Equation (6)) are given by Equations (14)(16):[68]

r 1A k1A  C A ; 14

r 2D k2D  C D  C OH
; 15

 
 
68600:168
r 2D 1:48  106  3600  exp
8:314  T
  
 
68035:058
2:68  108  3600  exp  C OH
 C D ;
Figure 1. Tubular reactor for synthesis of propylene oxide in Poliolchem
8:314  T
d.o.o. Tuzla, Bosnia and Herzegovina. 16

VOLUME 93, JANUARY 2015 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 79

 Table 1. Notation and molar masses of components The overall molar ow of the reaction mixture at the entrance to
reactor is given by the following equation:
Notation Component Molar mass [kg/kmol]

A propylene chlorohydrin, C3H7OCl 94.54 F T0 F A0 F B0 F C0 F F0 : 19

B calcium hydroxide, Ca(OH)2 74.1
C hydrochloric acid, HCl 36.45
D propylene oxide, C3H6O 58.08 Energy balance
E calcium chloride, CaCl2 110.9 The energy balance is described by the following equation:
F water, H2O 18
G propylene glycol, C3H8O2 76.09 dT U  a  T a
T
r 1A 
DH R1
r 2D 
DH R2 
:
dV P7
F i  C pi
i1
Because of the small solubility Ca(OH)2 in water, the concentra-
20
tion of OH
ions is constant and depends only on temperature:[6]
The overall heat transfer coefcient is calculated according to the
C OH
0:13
3:01  10
4  T: 17 equation:[12]
 1=4
For numerical simulation of the reactor performance, the kinetic T
Ta
U 1:22  : 21
models from three different sources were used.[68] These models D
are summarized in Table 2. It should be noted that kinetic models
for Equations (5) and (6) from three different sources were Heat capacities
combined in order to explore more different cases. Therefore, it is The calculation of specic heat capacities of the components A, C,
possible that different combinations of kinetic models for the rst D, F, and G are represented by the general Equation (22):
and for the second reaction will give the better results than the
original kinetic models. C pi a b  T g  T 2 d  T 3 : 22
Stoichiometry
The concentrations of components that participate in the reactions The values of the coefcients of equation for calculation of the
are given by the general Equation (18): specic heat capacity (22) are shown in Table 3. The mean heat
capacities are given for the components B and E.
  The general form of the equation for calculating the specic heat
Fi
C i C T0  ; 18 capacity of the rst reaction (3) is given by the Equation (23), while
F T0
its nal form is given by the Equation (24):

where i 1,7 (A, B, C, D, E, F, G). DC p1 C pD C pE 2C pF
C pA
C pB
C pC ; 23

Table 2. Applied kinetic models

Rate constant for the second reaction, k2D [m3/(kmol h)], or

Case Rate constant for the first reaction, k1A [h
1] rate law for the second reaction, r2D [kmol/(m3 h)]
   
1
187131
68048:1
k1A 3:394  1024  3600  exp Kotora et al.[8] k2D 2:68  1011  3600  exp Carr et al.[6]
8:314  T 8:314  T
   
2
66402:6
68048:1
k1A 3:02  109  3600  exp Carr et al.[6] k2D 2:68  1011  3600  exp Carr et al.[6]
8:314  T 8:314  T
 
  
3
187131
68600:168
k1A 3:394  1024  3600  exp Kotora et al.[8] r 2D 1:48  106  3600  exp
8:314  T 8:314  T
  
 
68035:058
2:68  108  3600  exp  C OH  C D
8:314  T
Carr et al.[6]
 
  
4
66402:6
68600:168
k1A 3:02  109  3600  exp Carr et al.[6] r 2D 1:48  106  3600  exp
8:314  T 8:314  T
  
 
68035:058
2:68  108  3600  exp  C OH  C D
8:314  T
Carr et al.[6]
   
5
187131
75362
k1A 3:394  1024  3600  exp Kotora et al.[8] k2D 4:7111  1012  3600  exp Molnar et al.[7]
8:314  T 8:314  T
   
6
66402:6
75362
k1A 3:02  109  3600  exp Carr et al.[6] k2D 4:7111  1012  3600  exp Molnar et al.[7]
8:314  T 8:314  T

80 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 93, JANUARY 2015

 Table 3. Coefcients of equation for calculation of molar heat capacities of components or mean molar heat capacities of components

Coefficients of equation for calculation the molar heat capacity

Mean heat capacity
Component a b g d Cp [kJ/(kmol K)]

A propylene chlorohydrins, 202.212 9.2336  10
1 2.9147  10
3 0

Zabransky and Ruicka[13]
B calcium hydroxide,[14] 87.5
C hydrochloric acid, ef and Oluji
c[15] 17.723 9.043  10
1
5.645  10
3 1.134  10
5
D propylene oxide,[16] 113.08
1.5085  10
1 6.728385  10
4 0
E calcium chloride,[17] 72.9
F water, ef and Oluji
c[15] 18.296 4.721  10
1
1.339  10
3 1.314  10
6
G propylene glycol, Perry and Green[18] 58.08 4.452  10
1 0 0

The overall yield of propylene oxide is calculated by the

DC p1
84:863 8:1241  10
1  T 7:251  10
4  T 2 Equation (32):

8:712  10
6  T 3 :
~D FD
24 Y : 32
F A0
F A
The general form of the equation for calculating the specic heat
capacity of the second reaction (4) is given by the Equation (25), The degree of conversion of propylene chlorohydrin is calculated
while its nal form is the Equation (26): by the Equation (33):

DC p2 C pG
C pD
C pF ; 25 F A0
F A
XA : 33
F A0
DC p2
73:296 1:2395  10
1  T 6:661615  10
4  T 2

1:314  10
6  T 3 : Solutions of the model equations
26 The mathematical model is consisted of eight rstorder ordinary
differential equations (seven mole balance equations and one heat
Heat of reactions
balance equation) and several corresponding nonlinear algebraic
The heat of the rst reaction (5) is calculated according to the equations. The solutions give the molar ow rates of the
Equation (27), while its nal form is the Equation (28): components and temperature of the reaction mixture.
Since an analytical solution of such system of differential
ZT equations is not possible, a numerical solution will be used. As a
DH R1 DH 0R1 DC p1 dT; 27 numerical method for solving, the RungeKuttaFehlberg (RKF45)
T ref method was chosen.
The Polymath Educational Version 6.0[19] is used for solution of the
DH R1
27522:294
84:863  T 4:062  10
1  T 2 system of differential equations and nonlinear algebraic equations.
28 Microsoft Ofce Excel is used for the graphical presentation of the
2:417  10
4  T 3
2:178  10
6  T 4 : simulation results.
Input data required for the implementation of mathematical
The heat of the second reaction (6) is calculated according to the model are reported in Table 4. It should be noted that three different
Equation (29), while its nal form is the Equation (30): data sets (marked with superscripts: a, b, c) from an industrial
tubular reactor were provided for the purpose of this study.
ZT Different data sets will be used for comparison with simulation
DH R2 DH 0R2 DC p2 dT; 29 results in order to nd the most appropriate kinetic model as well as
to validate the model. Anyway, model validation based on three
T ref
runs will be more reliable than model validation based on a
single run.
DH R2
70238:852
73:296  T 6:197  10
2  T 2
30
2:22503  10
4  T 3
3:28  10
7  T 4 : RESULTS AND DISCUSSION

Comparison of The Used Kinetic Models and Actual Process Data

Selectivity, yield, and conversion
The overall selectivity of propylene oxide related to propylene
glycol is calculated by the Equation (31):
~DG F D :
S 31
FG
The results of numerical simulations as well as measured (derived)
values from Poliolchem Inc. Tuzla[20] are shown in Table 5. The
rst three lines of Table 5 correspond to three different industrial
runs and the cases from 1 to 6 are the corresponding simulations
(the same for Table 6). The percentage deviations of these values
are reported in Table 6.

VOLUME 93, JANUARY 2015 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 81

 Table 4. Parameters used in the mathematical model

Parameter Numerical value Unit

Volume of reactor, V 2.039 m3

Surface of heat exchange, A 40.78 m2
Diameter of reactor, D 0.20 m
Reactor length, L 64.9 m
Entering molar flow of propylene chlorohydrin, FA0 44.789a kmol/h
39.565b kmol/h
35.880c kmol/h
Entering molar flow of calcium hydroxide, FB0 49.739a kmol/h
43.783b kmol/h
41.060c kmol/h
Entering molar flow of hydrochloric acid, FC0 46.919a kmol/h
41.455b kmol/h
37.340c kmol/h
Entering molar flow of water, FF0 6278.633a kmol/h
5700.780b kmol/h
5000.66c kmol/h
Ambient temperature, Ta 293.15a K
287.60b K
290.30c K
Temperature of the reaction mixture at the entrance to reactor, T0 333.15a K
328.70b K
330.35c K
Concentration of the reaction mixture at the entrance to reactor, CT0 52.840a kmol/m3
49.34b kmol/m3
50.55c kmol/m3
a
real data set 1.
b
real data set 2.
c
real data set 3.

As it can be seen from Tables 5 and 6, the two cases with the
smallest percentage deviations from the actual values are cases 2
and 6. By comparing cases 2 and 6, it can be seen that case 6 has
smaller deviation of the molar ow rate of propylene oxide
(component D), while case 2 has smaller deviations of the three
parameters: the molar ow rate of propylene glycol (component
G), the overall selectivity of propylene oxide related to propylene
glycol as well as the overall yield of propylene oxide. For other
parameter values the deviations are equal or almost the same.
Therefore, case 2 represents the best choice of kinetic model for the
purpose of numerical simulations.
The two cases with the largest percentage deviations from the
actual values are cases 3 and 5. Case 5 has larger deviation of the
three parameters: the molar ow rate of propylene glycol
(component G), the overall selectivity of propylene oxide related
to propylene glycol and the overall yield of propylene oxide.
Therefore, case 3 represents the worst choice of kinetic model for
purpose of numerical simulations.
The measurements from an industrial reactor (results from
Table 5) showed that hydrochloric acid is not present in output
ow due to complete neutralization with a small excess of calcium
hydroxide. On the other hand, the results of numerical simulations
showed that hydrochloric acid is present in output ow. This
difference is due to fact that the neutralization of hydrochloric acid
is carried rapidly. Therefore, the model could not predict accurately
the changes of molar ow rates of hydrochloric acid as for other
process variables.
The comparison of the values of the rate constants for the
reaction to obtain propylene oxide in cases 2 and 3 showed a big
difference (even 10 000 times), while the comparison of the value
of the reaction rate of obtaining propylene glycol showed an even
greater difference (1 000 000 times). These signicant differences
of the simulation results for cases 2 and 3 can be explained by the
activation energy for the rst reaction. Namely, the activation
energy in case 3 is 2.8 times higher than that in the case 2. As a
consequence, signicantly lower conversion was achieved for the
approximately same temperature. On the other hand, the kinetic
model for the second reaction (cases 2 and 3) gives a small
contribution to the difference between the simulation results. It can
be seen from the comparison of simulation results for cases 2 and 4,
which have the same kinetic model for the rst reaction but
different kinetic models for the second reaction. For these cases,
simulation results are the same or similar for the most of simulated
values, except for the molar ows of propylene oxide and
propylene glycol. If the pH is higher than 12, the effective kinetic
constant for the noncatalyzed hydrolysis becomes negligible in
comparison to the effective kinetic for the acidbase catalyzed
hydrolysis reaction.[6] Therefore, the simplied kinetic model for
the second reaction in case 2 is important for industrial purposes
because the pH of solution in the reactor is maintained above 12.
The kinetic model for the second reaction in cases 5 and 6 has a low
inuence on the simulation results (when compared to the kinetic
model in cases 1 and 2) due to very close values of the frequency
factors and the activation energies. The ndings from this study
indicate the importance of the application appropriate kinetic data
for reactor analysis and optimization.
Numerical Simulation with The Adopted Kinetic Model
Figures 24 graphically show the proles of different parameters
(case 2) as a function of the reactor length. It can be observed that
the proles of parameters for different runs are similar, with
specic shifts due to their different initial values.

82 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 93, JANUARY 2015

 Table 5. Comparison of simulated and measured (derived) values

Case FA [kmol/h] FB [kmol/h] FC [kmol/h] FD [kmol/h] FE [kmol/h] FF [kmol/h] FG [kmol/h] T [K] S~DG Y~ D XA

Measured (derived) values 0.047a 4.039 0 43.865 45.803 6370.211 0.442 336.65 99.041 0.989 0.9900
0.250b 3.650 0 38.803 40.567 5790.450 0.281 332.90 138.089 0.987 0.9940
0.038c 4.287 0 35.110 36.630 5095.340 0.353 334.80 100.595 0.980 0.9990
1 44.756d 49.697 46.877 0.042 0.042 6278.716 0.0002 333.15 193.845 0.995 0.0009
39.550e 43.768 41.440 0.015 0.015 5700.81 6.03  10
5 328.70 251.552 0.996 0.0004
35.857f 41.037 37.317 0.023 0.023 5000.70 0.0001 330.35 193.376 0.995 0.0006
2 0.012d 4.953 2.133 44.326 44.786 6367.745 0.460 335.85 96.387 0.990 0.9997
0.091e 4.309 1.981 39.177 39.474 5779.431 0.296 331.28 132.190 0.992 0.9977
0.011f 5.183 1.463 35.507 35.877 5072.44 0.370 333.05 95.050 0.990 0.9997
3 44.756d 49.697 46.877 0.042 0.042 6278.717 3.302  10
5 333.15 1271.96 1.001 0.0009
39.550e 43.768 41.440 0.015 0.015 5700.81 8.843  10
6 328.70 1721.70 0.999 0.0004
35.857f 41.037 37.317 0.023 0.023 5000.70 1.742  10
5 330.35 1301.63 0.999 0.0006
4 0.012d 4.953 2.133 44.858 44.786 6368.277 0.073 335.74 618.395 1.002 0.9997
0.092e 4.310 1.982 39.428 39.473 5779.68 0.044 331.22 887.006 0.999 0.9977
0.011f 5.191 1.471 35.812 35.869 5072.34 0.056 332.97 633.60 0.998 0.9997
5 44.756d 49.697 46.877 0.042 0.042 6278.716 0.0003 333.15 154.581 0.994 0.0009
39.550e 43.768 41.440 0.015 0.015 5700.81 7.293  10
5 328.70 207.893 0.995 0.0004
35.857f 41.037 37.317 0.023 0.023 5000.70 0.0001 330.35 157.67 0.994 0.0006
6 0.012d 4.953 2.133 44.199 44.786 6367.618 0.586 335.87 75.229 0.987 0.9997
0.091e 4.309 1.981 39.109 39.474 5779.36 0.365 331.29 107.111 0.991 0.9977
0.011f 5.191 1.471 35.407 35.869 5071.936 0.462 333.07 76.657 0.987 0.9997
The superscripts are shown only in the second column of Table 5, but these superscripts are the same in one particular row for all columns.
a
real data set 1.
b
real data set 2.
c
real data set 3.
d
simulation data set 1.
e
simulation data set 2.
f
simulation data set 3.

Figure 4 shows that the temperature of the reaction mixture
within the reactor rapidly increase at the reactor length of 40 m,
and then maintain almost constant value until the end of the
process. Values of the process parameters with the reactor length
are shown in Table 7. It is clear that increase of the reactor length
led to decrease of the overall selectivity of propylene oxide related
to propylene glycol and the overall yield of propylene oxide, and to
increase of the conversion of propylene chlorohydrin. The results
showed that for series of consecutive reactions in this study
(Equations (5) and (6)), the most important variable is spacetime
(volume of reactor/entering volumetric ow rate) because space
time is directly proportional to the reactor length. The rst reaction

Table 6. Deviations of simulated values from measured (derived) values [%]

Case FA FB FC FD FE FF FG T S~DG Y~ D XA

1 95125.53d 1130.40
99.91
99.91
1.44
99.95
1.04 95.72 0.61
99.91
15720.00e 1099.12
99.96
99.96
1.55
99.98
1.26 82.17 0.91
99.96
94260.53f 857.24
99.93
99.94
1.86
99.97
1.33 92.23 1.53
99.94
2
74.47d 22.62 1.05
2.22
0.04 3.83
0.24
2.68 0.07 0.98

63.60e 18.05 0.96
2.69
0.19 5.33
0.49
4.27 0.56 0.37

71.05f 20.90 1.13
2.06
0.45 4.82
0.52
5.51 1.02 0.07
3 95125.53d 1130.40
99.91
99.91
1.44
100.01
1.04 1184.28 1.21
99.91
15720.00e 1099.12
99.96
99.96
1.55
100.00
1.26 1146.80 1.26
99.96
94260.53f 857.24
99.93
99.94
1.86
82.58
1.33 1193.93 1.96
99.94
4
74.47d 22.63 2.26
2.22
0.03
116.38
0.27 524.38 1.31 0.98

63.20e 18.08 1.61
2.70
0.19
84.34
0.50 542.34 1.22 0.37

71.05f 21.09 2.00
2.08
0.45
84.14
0.55 529.85 1.84 0.07
5 95125.53d 1130.40
99.90
99.91
1.44
99.94
1.04 56.08 0.46
99.91
15720.00e 1099.12
99.96
99.96
1.55
99.97
1.26 50.58 0.81
99.96
94260.53f 857.24
99.93
99.94
1.86
99.97
1.33 56.74 1.43
99.94
6
74.47d 22.62 0.76
2.22
0.04 32.25
0.23
24.04
0.21 0.98

63.60e 18.05 0.79
2.69
0.19 29.89
0.48
22.43 0.41 0.37

71.05f 21.09 0.85
2.08
0.46 30.88
0.52
23.80 0.71 0.07
The superscripts are shown only in the second column of Table 6, but these superscripts are the same in one particular row for all columns.
d
simulation data set 1.
e
simulation data set 2.
f
simulation data set 3.

VOLUME 93, JANUARY 2015 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 83

Figure 2. Molar ow rates of propylene chlorohydrin, calcium hydroxide, hydrochloric acid, propylene oxide, calcium chloride and propylene glycol as a
function of reactor length: (a) run 1, (b) run 2, (c) run 3.
(Equation (5)), formation of desired product D) is faster than the
second reaction (Equation (6)), formation of undesired product G),
but if the reactor is too long, the desired product D will be converted
to the undesired product G. Therefore, the molar ow rate of
desired product (propylene oxide) is decreased, and the molar ow
rate of undesired product (propylene glycol) with the increase of
the reactor length.
The results of numerical simulations for the case of different
temperatures of input ow in reactor are reported in Table 8. The
overall selectivity of propylene oxide related to propylene glycol is
strongly affected by the temperature of input ow. Changing the
temperature from 313.15 K to 353.15 K, the overall selectivity is
decreased for even 1977.4 %. With the increase of the input ow
temperature, the overall yield of propylene oxide is decreased for
3.4 %, while degree of conversion is increased for 22.7 %. The
Figure 3. Molar ow rate of water as a function of reactor length.
84 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING
molar ow rate of propylene glycol is increased for 18.5 % with
increase of the input ow temperature. For the same temperature
interval, the molar ow rate of undesired product (propylene
glycol) is decreased for even 2367.7 %. These results showed that
molar ow rate of propylene glycol is the most sensitive variable
to temperature changes, followed by the overall selectivity of
propylene oxide related to propylene glycol. Since the activation
energy of the rst reaction is lower than the activation energy of
the second reaction, reactions should be carried out at the lower
temperatures to maximize the overall selectivity of propylene
oxide related to propylene glycol. This is consistent with the
observation of Levenspiel,[21] who gave a general rule about low
temperature being more favourable for the reaction with lower
activation energy. However, dealing with too low temperatures
Figure 4. Temperature of the reaction mixture as a function of reactor
length.
VOLUME 93, JANUARY 2015
 Table 7. Inuence of reactor length on molar ow rates of components, output ow temperature, overall selectivity of PO related to PG, overall yield PO
and conversion of PC (POpropylene oxide, PG propylene glycol, PC propylene chlorohydrin)

L [m] FA [kmol/h] FB [kmol/h] FC [kmol/h] FD [kmol/h] FE [kmol/h] FF [kmol/h] FG [kmol/h] T [K] S~DG Y~ D XA
d
54.9 0.044 4.985 2.165 44.38 44.75 6367.76 0.378 335.8 117.5 0.9916 0.9990
0.239e 4.457 2.129 39.085 39.326 5779.19 0.241 331.3 162.5 0.9939 0.9939
0.039f 5.211 1.491 35.545 35.849 5072.05 0.303 333.0 117.1 0.9915 0.9989
64.9 0.012d 4.953 2.133 44.33 44.79 6367.74 0.460 335.8 96.4 0.9897 0.9997
0.091e 4.309 1.981 39.177 39.474 5779.43 0.296 331.3 132.2 0.9924 0.9977
0.011f 5.183 1.463 35.507 35.877 5072.04 0.370 333.0 96.1 0.9897 0.9997
74.9 0.003d 4.944 2.124 44.25 44.80 6367.68 0.542 335.9 81.7 0.9879 0.9999
0.035e 4.253 1.925 39.178 39.530 5779.49 0.353 331.3 111.2 0.9911 0.9991
0.003f 5.175 1.455 35.449 35.885 5071.99 0.436 333.1 81.3 0.9878 0.9999
The superscripts are shown only in the second column of Table 7, but these superscripts are the same in one particular row for all columns.
d
simulation data set 1.
e
simulation data set 2.
f
simulation data set 3.

may lead that the desired reaction does not proceed to any
signicant extent.[22] The maximum molar ow rate of propylene
oxide was achieved with temperature of the entering reaction
mixture of 57 8C (330.15 K) (see Table 8). Optimization problem
would be reduced by nding the temperature prole along the
length of reactor in which the maximum yield of propylene oxide
can be achieved.
The results of numerical simulations for the case of different
molar ow of components are reported in Table 9. No signicant
changes on the overall yield of propylene oxide, the degree of

Table 8. Inuence of input ow temperature on overall selectivity of PO related to PG, overall yield of PO, conversion of PC and molar ow rates of PO and
PO (POpropylene oxide, PG propylene glycol, PC propylene chlorohydrin)

T0 [K] S~DG Y~ D XA FD [kmol/h] FG [kmol/h]

d
353.15 26.935 0.964 1.0000 43.194 1.604
26.223e 0.963 1.0000 38.111 1.453
22.156f 0.957 1.0000 34.338 1.550
348.15 35.870d 0.973 1.0000 43.583 1.215
34.916e 0.972 1.0000 38.463 1.102
29.467f 0.967 1.0000 34.710 1.178
343.15 48.670d 0.980 0.9999 43.896 0.902
47.357e 0.979 1.0000 38.747 0.818
39.875f 0.976 1.0000 35.010 0.878
338.15 67.531d 0.985 0.9999 44.144 0.654
65.662e 0.985 1.0000 38.971 0.593
55.062f 0.982 1.0000 35.248 0.640
333.15 96.387d 0.989 0.9997 44.326 0.460
93.606e 0.989 0.9998 39.138 0.418
77.973f 0.987 0.9999 35.432 0.454
330.15 121.348d 0.992 0.9986 44.371 0.366
117.729e 0.992 0.9988 39.187 0.333
97.534f 0.990 0.9997 35.511 0.364
328.15 142.603d 0.993 0.9966 44.334 0.311
138.251e 0.993 0.9970 39.165 0.283
114.075f 0.991 0.9989 35.539 0.312
323.15 219.392d 0.995 0.9788 43.648 0.199
212.352e 0.995 0.9808 38.625 0.182
173.633f 0.994 0.9906 35.347 0.204
318.15 347.957d 0.997 0.9235 41.252 0.118
336.536e 0.997 0.9287 36.634 0.109
274.290f 0.996 0.9558 34.178 0.125
313.15 559.544d 0.998 0.8149 36.443 0.065
541.311e 0.998 0.8235 32.521 0.060
443.131f 0.998 0.8712 31.195 0.070
The superscripts are shown only in the second column of Table 8, but these superscripts are the same in one particular row for all columns.
d
simulation data set 1.
e
simulation data set 2.
f
simulation data set 3.

VOLUME 93, JANUARY 2015 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 85

 Table 9. Inuence of entering molar ows of components on overall selectivity of PO related to PG, overall yield of PO, conversion of PC and molar ow
rates of PO and PO (PO propylene oxide, PG propylene glycol, PC propylene chlorohydrin)

Percentage increase or decrease of the

entering molar flow of components [%] S~DG Y~ D XA FD [kmol/h] FG [kmol/h]

20 119.591d 0.992 0.9989 53.254 0.445

165.548e 0.994 0.9934 46.884 0.283
119.256f 0.992 0.9988 42.656 0.358
10 107.817d 0.990 0.9994 48.797 0.453
148.590e 0.993 0.9959 43.054 0.290
107.482f 0.991 0.9993 39.079 0.364
0 96.387d 0.989 0.9997 44.326 0.460
132.2e 0.9924 0.9977 39.177 0.296
96.1f 0.9897 0.9997 35.507 0.370

10 85.299d 0.988 0.9998 39.847 0.467
116.372e 0.991 0.9988 35.264 0.303
84.972f 0.988 0.9999 31.912 0.376

20 74.546d 0.987 0.9999 35.363 0.474
101.148e 0.990 0.9995 31.327 0.310
74.233f 0.987 1.0000 28.321 0.381
The superscripts are shown only in the second column of Table 9, but these superscripts are the same in one particular row for all columns.
d
simulation data set 1.
e
simulation data set 2.
f
simulation data set 3.

conversion and the molar ow rate of propylene glycol are CONCLUSIONS

observed when molar ows of components are changed. Con-
versely, a marked effect is evidenced in the overall selectivity of
propylene oxide related to propylene glycol and the molar ow of
propylene oxide. The increases of all input ows for 20 % from
default values lead to: the increase of the overall selectivity of PO
related to PG for 24 %, the increase of overall yield of PO for 0.3 %,
the decrease of conversion of PC for 0.1 %, the increase of the molar
ow rate of PO for 20 % and the decrease of molar ow rate of PG
for 3.4 %. The decreases of all input ows for 20 % from default
values lead to: the decrease of the overall selectivity of PO related to
PG for 29.3 %, the decrease of overall yield of PO for 0.2 %, the
increase of conversion of PC for 0.02 %, the decrease of the molar
ow rate of PO for 25.3 % and the increase of molar ow rate of
PG for 3 %. These results showed that overall selectivity of
propylene oxide related to propylene glycol is the most sensitive
variable to input ow changes, followed by the molar ow rate of
propylene oxide. However, the relationship between molar ows
of components and input ow temperature was not taken into
account in this study.
The developed mathematical model can be used to improve the
understanding of the process, to guide the choice of the optimal
operation conditions, and to suggest possible changes to the reactor
design in order to improve the process performance. In order to get
the simulation results of higher accuracy, the radial gradients in the
reactor should be studied. In addition to inuences of the reactor
length, the input ow temperature and molar ows of components
on process performance, some of the economic indicators of the
process should also be taken into account, such as prices of
reactants and products, prices of reactor construction materials, the
costs of separating the components of the product mixture by
distillation, etc. Moreover, a rigorous optimization has not been
carried in this study, and therefore it will be the objective of the
next paper. For the purpose of considering the radial gradients as
well as optimization, additional data sets from an industrial reactor
should be provided.
This paper compared different kinetic models for the synthesis of
propylene oxide by chlorohydrin process using real data from an
industrial tubular reactor. In order to accomplish this objective, a
simplied reactor model was developed and applied on the kinetic
models. The ability of the model to simulate the reactor behaviour is
highly dependent on kinetics. The results of numerical simulations
showed good agreement with actual process values. The inuences
of the reactor length, the input ow temperature and the entering
molar ows of components on the process performance were
studied using the validated mathematical model. The maximum
molar ow rate of propylene oxide was achieved with temperature
of the entering reaction mixture of 57 8C. The increase of the input
ows by 20 % leads to increase of the overall selectivity for 24 %.
The reactor length, the input ow temperature and the molar ows
of components have the strongest inuence on overall selectivity of
propylene oxide related to propylene glycol. A rigorous optimiza-
tion will be the objective of the future study.
NOMENCLATURE
a area of heat exchange per unit volume of reactor [m
1]
A surface of heat exchange [m2]
CA concentration of propylene chlorohydrin [kmol/m3]
CB concentration of calcium hydroxide [kmol/m3]
CC concentration of hydrochloric acid [kmol/m3]
CD concentration of propylene oxide [kmol/m3]
CE concentration of calcium chloride [kmol/m3]
CF concentration of water [kmol/m3]
CG concentration of propylene glycol [kmol/m3]
C OH
concentration of OH ions [kmol/m3]
CpA specific heat capacity of propylene chlorohydrin
[kJ/(kmol K)]
CpB specific heat capacity of calcium hydroxide [kJ/(kmol K)]

86 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING VOLUME 93, JANUARY 2015

CpC specific heat capacity of hydrochloric acid [kJ/(kmol K)] [7] A. Molnar, J. Marko, L. Jelemensky, Chem. Pap. 2002, 56,
CpD specific heat capacity of propylene oxide [kJ/(kmol K)] 357.
CpE specific heat capacity of calcium chloride [kJ/(kmol K)] [8] M. Kotora, Z. vandova, J. Marko, L. Jelemensky, Pet. Coal
CpF specific heat capacity of water [kJ/(kmol K)] 2005, 47, 26.
CpG specific heat capacity of propylene glycol [kJ/(kmol K) [9] S. Carr, M. Morbidelli, E. Santacesaria, G. Buzzi, Chem. Eng.
CT0 concentration of the reaction mixture at the entrance to Sci. 1979, 34, 1133.
reactor [kmol/m3]
[10] S. Elgue, J. M. Le Lann, M. Cabassud, L. Prat, J. Cezerac,
D diameter of reactor [m]
Chem. Biochem. Eng. 2003, 17, 43.
E activation energy [kJ/kmol]
FA molar flow rate of propylene chlorohydrin [kmol/h] [11] Poliolchem Inc. Tuzla, Annual reports from facilities, Tuzla
FB molar flow rate of calcium hydroxide [kmol/h] (19762010).
FC molar flow rate of hydrochloric acid [kmol/h] [12] R. B. Bird, W. E. Stewart, E. D. Lightfoot, Transport
FD molar flow rate of propylene oxide [kmol/h] Phenomena, 2nd edition, John Wiley and Sons, Inc., New
FE molar flow rate of calcium chloride [kmol/h] York 2002.
FF molar flow rate of water [kmol/h] [13] M. Zabransky, V. Ruicka Jr., J. Phys. Chem. Ref. Data 2004,
FG molar flow rate of propylene glycol [kmol/h] 33, 1071.
FA0 entering molar flow of propylene chlorohydrin [kmol/h] [14] Calcium hydroxide (data page), Wikipedia, accessed 5
FB0 entering molar flow of calcium hydroxide [kmol/h] January 2014. http://en.wikipedia.org/wiki/Calcium_hydr-
FC0 entering molar flow of hydrochloric acid [kmol/h] oxide_(data_page).
FF0 entering molar flow of water [kmol/h]
[15] F. ef, . Oluji
c, Projektiranje procesnih postrojenja, SKTH/
FT0 overall molar flow of the reaction mixture at the entrance
Kemija u Industriji. (in Croatian), Zagreb 1988.
to reactor [kmol/h]
DH R1 heat of the first reaction [kJ/kmol] [16] D. L. Trent, Propylene Oxide, KirkOthmer Encyclopedia of
DH R2 heat of the second reaction [kJ/kmol] Chemical Technology, John Wiley & Sons, Inc., 2001, accessed
k1A rate constant for the first reaction [h
1] 5 January 2014. www.scribd.com/doc/30134187/Propylene
k2D rate constant for the second reaction [m3/(kmol h)] Oxide.
L reactor length [m] [17] Calcium chloride (data page), Wikipedia, accessed 5
r1A rate of the first reaction [kmol/(m3 h)] January 2014. http://en.wikipedia.org/wiki/Calcium_chlor-
r2D rate of the second reaction [kmol/(m3 h)] ide_(data_page).
~DG
S overall selectivity of propylene oxide related to propylene [18] R. H. Perry, D. W. Green, Perrys Chemical Engineers
glycol Handbook, 7th Edition, McGrawHill, New York 1997.
T temperature of the reaction mixture [K] [19] M. Shacham, M. B. Cutlip, M. Elly, Polymath Educational
Ta ambient temperature [K] Version 6.0, Control Data Corporation, 2004.
T0 temperature of the reaction mixture at the entrance to [20] Poliolchem Inc. Tuzla, List of parameters of saponication
reactor [K] and distillation, Tuzla (19762010).
U overall heat transfer coefficient [kJ/(h K)]
[21] O. Levenspiel, Chemical Reaction Engineering, 3rd edition,
V volume of reactor [m3]
John Wiley & Sons, Inc., New York 1999.
XA degree of conversion for propylene chlorohydrin
~D
Y overall yield of propylene oxide [22] H. S. Fogler, Elements of Chemical Reaction Engineering, 4th
edition, PrenticeHall Inc., Englewood Cliffs, USA 2006.

ACKNOWLEDGEMENTS
Manuscript received February 26, 2014; revised manuscript received
The authors are grateful to the people from the company March 26, 2014; accepted for publication April 21, 2014.
Poliolchem d.o.o. Tuzla for providing technical data, especially
Mr. Ago Nadarevi
c, process engineer.

REFERENCES
[1] J. R. Valbert, J. G. Zajacek, D. I. Orenbuch, Propene oxide
and glycol, Encyclopedia of Chemical Processing and Design,
Marcel Dekker, New York 1993, p. 88.
[2] A. Seubsai, M. Kahn, S. Senkan, ChemCatChem 2011, 3, 174.
[3] D. Kalich, U. Wiechern, J. Linder, Propylene oxide,
Encyclopedia of Industrial Chemistry, 5th Edition, Verlag
Chemie Weinheim, Germany Vol. A22, 1993, p. 239.
[4] T. A. Nijhuis, M. Makkee, J. A. Moulijn, B. M. Weckhuysen,
Ind. Eng. Chem. Res. 2006, 45, 3447.
[5] F. Bezzo, A. Bertucco, A. Forlin, M. Barolo, Sep. Purif.
Technol. 1999, 16, 251.
[6] S. Carr, E. Santacesaria, M. Morbidelli, L. Cavalli, Chem.
Eng. Sci. 1979, 34, 1123.

VOLUME 93, JANUARY 2015 THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING 87