Korean J. Chem. Eng.

, 26(6), 1504-1511 (2009)

DOI: 10.1007/s11814-009-0254-z
RAPID COMMUNICATION

Simulation and optimization of ethanol amine production plant

Gholamreza Zahedi*,, Saeideh Amraei**, and Mazda Biglari***

*Process Systems Engineering Centre (PROSPECT), Faculty of Chemical and Natural Resources Engineering,
Universiti Teknologi Malaysia, UTM Skudai, 81310 Johor Bahru, Johor, Malaysia
**Simulation and Artificial Intelligence Research Center, Chemical Engineering, Department, Razi University, Iran
***Department of Chemical Engineering, Ryerson University, Toronto, Ontario, Canada M5B 2K3
(Received 6 April 2009 accepted 7 April 2009)

AbstractAn industrial Ethanol Amine (EA) production plant was simulated and optimized. Due to lack of accurate
reaction rate information, the first step involved obtaining reliable kinetic data from the SRI (Stanford Research In-
stitute) industrial database and calculation using error minimization method. In the next step, by implementing the ob-
tained reaction kinetics the whole plant was simulated using Hysys software. Simulation results were compared with
the SRI data and showed that there is acceptable agreement between simulation and the measured industrial data. In
the next step of study by applying the gradient search (GS) optimization technique the plant was optimized using: feeding
ammonia to ethylene oxide (EO) molar ratio, water flow rate in the feed stream, and reactor temperature as optimization
variables. Employing process profit as objective function the optimal operating conditions were found to be: ammonia
to EO ratio of 5 (mol/mol), water flow rate of 52.59 kg mol/hr and reactor temperature of 85 oC.
Key words: Simulation, Optimization, Ethanol Amine Production, Ethylene Oxide, Ammonia

INTRODUCTION sity and have ammonia-like odor. The EA freezing point is consid-
erably decreased by adding water. MEA and DEA are soluble, with
Regarding the broad application of Ethanol Amine (EA), its pro- any proportion, in water and alcohol but considerably less in ether.
duction is of great interest to the petrochemical industry. The indus- All amines form white crystalline solids when freeze [7].
trial production of EA was accomplished in 1930 followed by the Very few quantitative studies have been conducted for EA syn-
large-scale production in 1945, when alkoxylation with ethylene thesis. One source is the study conducted by Japanese researchers
oxide (EO) and propylene oxide was replaced by chlorohydrins [1,2]. [8], yet the lack of confirmed data renders it unsuitable to construct
The commercial importance of EA increased from 1970, when in- further modeling and simulation based on a few kinetic data.
dustrial production of EO began. The modern industrial production In Iran, EA is produced at Arak Petrochemical Complex (APC).
of EA is operated exclusively by reaction of EO and excess amount In this plant EA is produced from a direct reaction of EO and aqueous
of ammonia in presence of water [3]. It has been proved that a ammonia solution in liquid phase (without addition of catalyst) in a
small amount of water is necessary to promote reaction. When there tubular fixed bed reactor. EO is also produced in the plant, whereas
is not any water in the system, EO does not react with ammonia. the required ammonia is provided by the domestic petrochemical
In fact, water plays a catalytic role in the corresponding reactions industry. The annual APC production capacity of EA is 30,000 tons
[2,4]. Based on operational conditions monoethanolamine (MEA), consisting of 10,000 tons MEA, 8,000 tons TEA and the rest is DEA.
diethanolamine (DEA) and triethanolamine (TEA) are produced Purity of the produced MEA and DEA is 99%, while TEA is pro-
through three parallel-consecutive competitive reactions [1]. duced in 99% and 85% purities [6]. This paper aims to simulate
EAs react with a large number of compounds such as ammonia, and investigate optimal operating condition of the MEA plant. Based
carbon dioxide, sulfuric acid, sodium hydroxide, carbon disulfide, on our literature review no work has been done in this case regard-
thionyl chloride and also with acids and aromatic aldehydes [4-6]. less of the importance of MEA.
In general, they are very desirable as H2S and CO2 absorbents for
natural gas treatment [7]. Because of alcoholic characteristics, MEA EA PROCESS DESCRIPTION
and DEA have broader industrial applications. On the other hand,
DEA and TEA are used in making detergents, textiles, pharmaceu- The commonly utilized process of EA production is the SRI pro-
ticals, emulsifiers in drilling and excavation of oil wells, corrosion cess, which has been typically designed for producing 45,000 ton/
inhibitors as well as additives to cement [2]. year EA from ammonia and EO, the production capacity that most
EAs are hygroscopic, colorless and viscous liquids at room tem- of the USA, Western Europe, and Japanese factories fall in. The
perature. TEA boils at 27 oC, whereas the other two EAs have higher operational conditions of SRI process have been regulated to pro-
boiling points. Their densities are slightly greater than water den- hibit formation of any by-product. This designing concept, however,
is not employed in the industry, and the industrial applications dif-
To whom correspondence should be addressed.

fer from the SRI process in that the equipment and plant structure
E-mail: grzahedi@razi.ac.ir are different, but the reaction is identical [1].
1504
 Simulation and optimization of ethanol amine production plant 1505

Fig. 1. Simplified flow diagram of ethanol amine production plant (Adapted from [1]).

Fig. 1 depicts a simplified flow diagram of an EA-producing plant. duced) to 3 (more even distribution of the three products).
The flow diagram has two major sections: the reactor and the prod- Optimization of other equipment within the plant has been also
uct separation. EO is added to the concentrated ammonia solution suggested. For example, the distillation columns can be optimized;
and sent to the reactor section. In general, this section consists of four however, optimization of the EAs distillation columns is beyond the
series of reactors that are adiabatically controlled by an inner cool- scope of this work and may be found in other sources (for exam-
ant. The reactors each are 3 m in length and 0.47 m in diameter. In ple, [11] and references therein).
the reactor section, the 99 wt% ammonia solution reacts with the
EO. Almost all of the EO is converted in this section. EO must be EVALUATION OF REACTION RATE CONSTANTS
added to ammonia solution; otherwise, an explosive polymeric reac-
tion occurs. The output of the reactor enters the ammonia stripper There are some industrial processes for the EAs production, in-
for un-reacted ammonia to be separated as vapor phase and to be cluding:
sent, after cooling, to an ammonia absorption tower as the make- BASF Process
up ammonia. In the absorber, ammonia is separated and recycled Dow Process
to the feed. The bottom of the ammonia stripper is fed into a vaporiz- Union Carbide Process
ing system to separate water and send it to the absorber. The con- Euteco SPA Process
centrated dried EAs from the bottom of vaporizer enter into the suc- Nippon Shokubai Process
cessive vacuum distillation columns. In the distillation towers MEA Oxiten Process
(with purity of 99 wt%), DEA (98.5 wt% purity), commercial All of the above processes exploit reaction of ammonia with EO
TEA (85 wt% TEA and 15 wt% DEA) and TEA (with purity of but are different in the operational conditions (such as residence
99 wt%) are produced [1,2,4]. time, operational temperature and pressure, ammonia concentration
An EA production plant is very flexible in order to adapt to market and other feed specifications) and equipment (such as adiabatic or
trends and to produce the demanded EA [3]; i.e., many operational isothermal reactors and the number of reactors). Three reactions
parameters may be adjusted to optimize the system to produce de- have mainly been recognized in the ethyl amine production [3,8]:
sirable product.
One important operational variable affecting product quality is (1)
temperature [3]. Higher than required temperature in the ammonia
stripper changes color of the product (EAs) and lowers its quaility.
Other targetted optimization variables include water content of am- (2)
monia solution, and ammonia/EO feedstock ratio. The higher the
water feed to the system, the lower the operational pressure and tem-
perature would be, which in turn results in a higher product quality.
However, the water addition has to be limitted to decrease utility (3)
consumption. Also, a significant amount of water content interferes
with the reactions of interest in favor of a side reaction of water with
EO to produce ethylene glycol. All of the above reactions occur simultaneously in parallel and in
The ammonia/EO ratio determines the proportion of final prod- series. The ratio of each EA in the final product is controlled by con-
uct of EA type [8]. Although this ratio is important, we did not find trolling ammonia to EO ratio [8].
a systematically optimized value among the literature for this ratio. Kinetic information pertaining to the reaction of EO with ammo-
A non peer-reviewed estimation was identified [10] in which this nia is not conclusive. Reference [1] reports the effect of different
ratio was enveloped between 10 (when only MEA is virtually pro- input aqueous ammonia solution on the produced EA characteris-
Korean J. Chem. Eng.(Vol. 26, No. 6)
1506 G. Zahedi et al.

tics. The reaction of EO with dilute aqueous ammonia solution (less Other relations were presented for the specific rate of contribut-
than 3 wt% of NH3) at 15 oC and variable molar ratio of 0.5/1 to ing reactions as Eqs. (5) to (7) [8]:
20/1 of ammonia to EO were studied. The data were explained by
11000
k1 = (41+ 4[ H2 O] ) 10 exp ------------------
2 2
an Arrhenius-type relationship (Eq. (4)) with activation energy, and (5)
RT
pre-exponential constants for the specific rate of the reactions are
given in Table 1. k2=(7.20.042[H2O])k1 (6)

E k3=(160.22[H2O])k2 (7)
k = k0 exp ------- (4)
RT
Assuming a first order conversion of EO in the temperature range of
where E is activation energy and k0 is a pre-exponential constant. 20 oC to 125 oC and ammonia solution of 50 to 100 wt%, the general
equation has been proposed [12]:
7.6
C 14500
Table 1. Arrhenius parameters for Eq. (4) [1] k =1.3 107 --------- exp ------------------ (8)
100 RT
Reaction E k0
where C is the ammonia mass composition in the feeding solution.
EO+NH3 14.6 4.111080 Due to the uncertainty in the kinetics data, in this study the data
EO+MEA 15.7 1.131010 from reference [1] were used to optimize k-values (ko correspond-
EO+DEA 15.1 2.941090 ing to each specific rate) for the reactions.
EO+TEA 12.7 4.231060 Considering the reaction Eqs. of (1), (2) and (3), each compo-

Fig. 2. Optimization algorithm for EA kinetic parameter estimation.

November, 2009
 Simulation and optimization of ethanol amine production plant 1507

nent reaction rate can be described as follows [13]: vestigate the effects of different parameters on the process and sub-
d [NH3 ]
sequent optimization of process operation. For the process model-
------------------ = k1 [NH3 ][EO] (9) ing in this study Hysys 3.2 was used. Fig. 3 illustrates the process
dt
flowchart made in Hysys, with the four reactors are labeled as R101A
d [EO ] to R101D.
--------------- = k1 [NH3 ][ EO] k2[MEA ][EO] k3 [DEA ][EO] (10)
dt Appropriate equations were selected in order to estimate the phys-
d [MEA ] ical properties of all components in the system. The results of simu-
--------------------- = k1[NH3 ][EO] k2 [MEA][EO ] (11)
dt lation are strongly influenced by the selected Hysys thermodynamic
d [DEA ]
model, i.e., selection of an inappropriate thermodynamic model will
-------------------- = k2 [MEA][EO ] k3[DEA ][EO ] (12) result in error in model predictions. The thermodynamic models in the
dt
Hysys environment are classified into three main groups: 1- equations
d [TEA ] of state such as SRK; 2- equations of activity coefficient such as
------------------- = k3[DEA] [EO] (13)
dt NRTL; 3- special thermodynamic models such as the amine model
For the optimization, Matlab 7.6 with solver ODE23s [14] was used prepared for a system consisting of amines, acidic gas and water [15].
to solve the series of above ordinary differential equations (ODEs).
Fig. 2 illustrates the proposed algorithm for obtaining kinetic con- SIMULATION RESULTS
stants. The algorithm first estimates the parameters (k1, k2, and k3),
followed by solving the ODEs and checking the error between re- To build the process flow diagram in the Hysys environment,
sults and the experimental data. The process continues until con- the composition of materials, equation of state and reactions were
vergence to the experimental data is reached. specified. The reactor feed specifications are presented in Table 3.
Results of the optimization of reaction rate constants are pre- Table 4 provides a comparison between the simulation and pro-
sented in Table 2. cess (SRI) data. For comparison, the errors in the simulation are
also given. Maximum error appears to be less than 4%, validating
PROCESS MODELING the simulation approach.
For different towers Tables 5, 6 and 7 present the estimated pro-
The objective of mathematical modeling and simulation is to in-

Table 3. Characteristics of feed to the reactor

Table 2. Results of optimization algorithm for kinetic parameters Temperature (oC) 85
estimation
Pressure (KPa) 1.02104
Reaction E k0 Total molar flow (kg mol/hr) 687.83
EO+NH3 19660 1.581107 Molar flow of ammonia (kg mol/hr) 529.09
EO+MEA 19660 04.92108 Molar flow of EO (kg mol/hr) 105.82
EO+DEA 19660 005.2108 Molar flow water (kg mol/hr) 052.92

Fig. 3. Hysys process flowchart of the EA production plant.

Korean J. Chem. Eng.(Vol. 26, No. 6)
1508 G. Zahedi et al.

Table 4. The characteristics of the reactors effluent stream (mole An objective function that is to be minimized or maximized.
ratio of ammonia to EO equals 5/1) For instance, in a manufacturing process, the objective is to maxi-
SRI result Hysys result Percent of mize profit or minimize cost.
Compound A set of unknowns or variables that control the objective func-
(kg/hr) (kg/hr) relative error
tion value. In the manufacturing problem, the variables might include
Ammonia 7842 7864 0.28
the amounts of different resources used or the time spent on each
EO - - -
activity.
Water 0935 0935 0.00
A set of constraints that allow the unknowns to take on certain
MEA 2456 2365 3.69
values but exclude others. For the manufacturing problem, it is mean-
DEA 2037 1966 3.47
ingless to spend a negative amount of time on any activity, so all
TEA 1338 1388 3.72 the time variables are constrained to non-negative values [12].
Generally, in optimization of processes, many parameters like
duction of different compounds including EAs and compare them temperature, pressure, raw materials flow rate, operational equip-
with SRI unseen data. Some significant differences between calcu- ment size and the process economy can be considered. To maxi-
lated and SRI data are observed. The main reason may be the limita- mize production rate, the optimum values of these parameters should
tion of the software, particularly in the thermodynamic models avail- be obtained. For the EA production process in this case, as the sizes
able in Hysys. Again, the differences between the simulated and of apparatus were already designed, this parameter is considered
SRI data are mostly less than 10%, which seems acceptable. constant. The operational pressure was also assumed to be con-
stant. Temperature, raw materials flow rate (feed) such as ammo-
PROCESS OPTIMIZATION nia, EO and water were considered as optimization variables, and
the process simulation was carried out for different rates. EA pro-
Optimization problems are made up of three basic ingredients: duction and cost objective function can be set as the objective func-

Table 5. Comparison between simulation and SRI data (kg/hr)

Bottom product of flash tower Bottom product of ammonia tower Bottom product of drying tower
Component
SRI data Hysys result SRI result Hysys result SRI result Hysys result
Ammonia 470.0 0491.60 0024.53 0024.49 - -
EO - - - - - -
Water 934.0 934.0 0934.21 0933.90 0006.35 0006.03
MEA 2456.60 2366.33 2454.34 2366.33 2444.36 2354.10
DEA 2037.12 1966.43 2037.12 1966.43 2036.67 1966.43
TEA 1338.73 1479.61 1338.73 1479.61 1338.73 1388.91

Table 6. Comparison between simulation and SRI data (kg/hr)

Overhead product of MEA tower Bottom product of MEA tower Bottom product of DEA tower
Component
SRI result Hysys result SRI result Hysys result SRI result Hysys result
Ammonia - - - - - -
EO - - - - - -
Water - - - - - -
MEA 2426.22 2329.03 0018.14 0018.66 - -
DEA 0016.32 0015.29 2020.34 1966.42 001.36 000.96
TEA 0000.45 0000.51 1338.28 1479.56 468.50 615.80

Table 7. Comparison between simulation and SRI data (kg/hr)

Overhead product of DEA tower The side product DEA tower Overhead product of TEA tower
Component
SRI result Hysys result SRI result Hysys result SRI result Hysys result
Ammonia - - - - - -
EO - - - - - -
Water - - - - - -
MEA 0018.14 0015.17 - - - -
DEA 1867.96 1541.75 150.56 145.94 001.36 000.52
TEA 0006.35 0005.74 410.86 410.86 421.75 423.12

November, 2009
 Simulation and optimization of ethanol amine production plant 1509

Table 8. Price of the raw materials and products [18]

Compound Price ($/tonne)
Ammonia 025.25
EO 155.26
Water 018.00
MEA 177.40
DEA 193.95
TEA (%99) 190.92
TEA (%85) 187.19

tion in this part of study. As there was not any idea about a suitable
practical objective function, both of the objective functions were
examined. The GS method was first applied to optimize the molar
production of MEA as the objective function as follows:
Fig. 4. Variation of objective function with molar ratio of ammo-
nia to EO. Objective function (J)=MoleMEA (14)

Figs. 4 to 6 depict effect of feed specification (ammonia molar flow/

EO molar flow) on the content of three EA types as final products.
Increasing ammonia content in the feed increases the production
of MEA and decreases the DEA and TEA in the product: no optimal
point appeared in the above figures. As a result, the objective func-
tion was changed to the profit function as a realistic approach de-
fined below [13]:
Objective function (J)=[products price][raw materials price] (15)

J=[kgMEApriceMEA+kgDEApriceDEA+kgTEA(%85)priceTEA(%85)
J=+kgTEA(%99) priceTEA(%99)][kgNH3priceNH +kgEOpriceEO
3

J=+kgH OpriceH O].

2 2
(16)

Table 8 presents price of the raw materials and products [18].

The constraints exposed to the optimization process were selected
based on operationally practical conditions [3,7,10] and included
the ammonia to EO molar flow of 5 to 20, and temperature range
of 85 oC to 120 oC. Temperatures beyond this range results mainly
Fig. 5. Effect of ammonia to EO molar ratio on DEA production in occurrence of side reactions.
rate.
OPTIMIZATION RESULTS

Figs. 7, 8 and 9 exhibit the variations of profit function (Eq. (16))

with changes in three selected controlling factors. At the maximum
profit, the optimum values of the variables were found to be: am-
monia to EO molar ratio 5/1; water molar flow 52.92 kg mol/hr and
reaction temperature 85 oC. As observed in Fig. 7, the ratio ammo-
nia/EO=5/1 generates maximum profit which overlaps the maxi-
mum profit observed in Figs. 8 and 9.

CONCLUSION AND REMARKS

An EA production plant was simulated by using Hysys soft-

ware. The results of simulation revealed that the simulated data have
a good agreement with the SRI data. There are, however, some errors
in the predicted data, mainly because of the softwares weakness in
this study, and particularly due to the available Hysys thermody-
Fig. 6. Effect of ammonia to EO molar ratio on TEA production namic models.
rate. The simulation also necessitates using accurate reaction rates.
Korean J. Chem. Eng.(Vol. 26, No. 6)
1510 G. Zahedi et al.

Since there are not convincing studies on the EAs production in

the literature, a simple kinetics optimization was carried out and re-
sults were confirmed by comparison with the available experimen-
tal data.
Next step involved optimization of the process to evaluate the
effects of feed flow rates and input reactor temperature, while the
pressure and equipment size were deemed constant. The MEA yield
appeared as an improper objective function having monotonic trends
with the selected controlling variables; therefore, no feasible results
were found. Consequently, the process profit was set as the objec-
tive function. Results of optimization showed that the optimal oper-
ation points are the ammonia to EO molar ratio of 5, reaction tem-
perature of 85 oC and water flow rate of 52.92 kg mol/hr.

NOMENCLATURE

Fig. 7. Economical profit (objective function) in terms of feeding C : concentration [%wt]

molar ratio of Ammonia to EO. E : activation energy [kcal mole1]
J : objective function
k0 : frequency factor [L mol1 min1]
k1, k2, k3 : kinetics constants [L mol1 min1]
ODE : ordinary differential equation
R : gas constant [kcal mol1 K1]
T : temperature [K], or otherwise stated
t : time [min]

Greek Letter
: error

REFERENCES

1. T. McMillan, SRI International 193(6), 1 (1991).

2. H. Hammer, Ethanolamines and propanolamines, Ullmanns Ency-
clopedia of Industrial Chemistry, 6th Ed., New York, Wiley-VCH,
10, 1 (2003).
3. P. Fassler and A. Celeghin, A new approach to an established prod-
Fig. 8. Variation of objective function with change of water molar uct: Cost-efficient production of ethanolamines, Sulzer Technical
inflow. Review, Issue No. 3 (2008).
4. M. Frauenkron, Ethanolamines and propanolamines, Ullmanns
Encyclopedia of Industrial Chemistry, 7th Ed., New York, Wiley-
VCH, 12, 475 (2008).
5. A. L. Kohl and R. Nielsen, Gas purification, 5th Ed., Houston, Texas,
Gulf Publishing Company, 41 (1997).
6. J. G. Spight, Chemical and process design handbook, New York,
McGraw-Hill (2002).
7. K. Weissermel, H.-J. Arpe, C. R. Lindley and S. Hawkins, Oxida-
tion products of ethylene, Industrial Organic Chemistry, Wiley-VCH,
159 (2003).
8. M. Hatta, T. Ito, M. Miki and T. Okabe, Reaction of ethylene oxide
with ammonia, Yukagaku (Journal of Japan Oil Chemists Society),
15, 215 (1966).
9. Http://www.arpc.ir/Default.aspx?tabid=143 (2008).
10. C. Ruehl, C. Hou, P. Lee and L. Armstrong, Design of a system of
ethanolamine reactors, Course project CENG 403, Rice Univer-
sity, Houston, Texas (1997).
11. K. Kolmetz, W. K. Ng, P. W. Faessler, K. Senthil, Y. T. Lim, A. W.
Fig. 9. Effect of reactor temperature on the process profit. Slolev and T. M. Zygula, Hydrocarbon Processing, 84(5), 77 (2005).
November, 2009
 Simulation and optimization of ethanol amine production plant 1511

12. C. Ruehl, C. Hou, P. Lee and L. Armstrong, Design of a system of 16. T. F. Edgar, D. M. Himmelblau and L. S. Lasdon, Optimization of
ethanolamine reactor, CENG 403 (1997). chemical processes, 2th Ed., New York, McGraw-Hill (2001).
13. O. Levenspiel, Chemical reaction engineering, 3th Ed., New York, 17. M. E. Masoumi, S. M. Sadrameli, J. Towfighi and A. Niaei, Jour-
Wiley (1998). nal of Energy, 31, 516 (2006).
14. MATLAB R2008a, Http://www.mathwork.com (2008). 18. Http://www.petrochem-ir.net/esales-internetsales-prodprice-fa.html/
15. Aspen Technolgoy, Inc., Hysys user guide, Ten Canal park, Cam- (2008).
bridge, USA (2005).

Korean J. Chem. Eng.(Vol. 26, No. 6)