Advanced Amine Process Simulation Model

Development
Bill Pennycook, Simon Jackson & Iain Martin
Davy Process Technology
20 Eastbourne Terrace
London W2 6LE
Iain.martin@davyprotech.com
www.davyprotech.com

1 Introduction Fig 1.1 Methylamines Unit

Davy Process Technology (DPT) has acquired a range of amine production technologies that
have been licensed for many years. Following acquisition, DPT engineers critically reviewed
the processes and then updated the technologies using advanced development and design
techniques which are similar to ones that have been applied on other DPT licensable
technologies. Key to this approach was establishing an accurate simulation model and it
became apparent that additional data would be required to augment and confirm the available
literature studies.

This paper describes the development of practical techniques that were used to prepare an
accurate multi-component simulation models from first principles with particular emphasis on
the process for the production of methylamines. These processes include complex aqueous
and organic systems and involve dealing with numerous azeotropes which are truly multi-
component, in contrast to a binary split of key components in the presence of other more or
less volatile minor components. It soon became apparent that the accurate measurement of
vapour liquid equilibrium data was particularly challenging for these highly odorous, volatile
and corrosive materials and a wide range of techniques and specific laboratory equipment
were necessary to accomplish the target.

This practical approach enabled analysis of the results for accuracy and provided confidence
in the simulations being undertaken. In turn, it was possible to move the technology forward
using these models to more accurately predict the performance and optimise equipment
design with associated value creation.

DPT now has a comprehensive portfolio of highly cost-effective process technologies for the
production of the majority of the industrial amines and these are available for licence. Of
particular interest is the production of methylamines since this is the largest volume
alkylamine produced and also since it feeds in to a range of derivatives which DPT is also
capable of licensing.

2 Description of DPT Flowsheet

Methylamines are produced by reacting ammonia with methanol at an elevated temperature

and pressure in the presence of a heterogeneous catalyst. A mixture of three amines is
formed, monomethylamine (MMA), dimethylamine (DMA) and trimethylamine (TMA). They
are separated and purified by distillation. A simplified set of equilibrium chemical reactions
that take place are shown below:

CH3OH + NH3 ? CH3NH2 + H2O

(MMA)
CH3OH + CH3NH2 ? (CH3)2NH + H2O
(DMA)
CH3OH + (CH3)2NH (CH3)3N + H2O
? (TMA)
By-product formation and degradation of the products to light ends is minimised by the use of
optimised processing conditions and a state-of-the-art catalyst. By-products include CO, H2
and higher molecular weight organics.

Figure 2.1: Schematic showing a DPT Methylamines Unit

Fresh methanol and ammonia are combined with recycled amines and ammonia and then
vaporised, superheated and fed to the catalytic methylamines converter. In the converter,
ammonia and methanol undergo an exothermic reaction to form mono, di- and tri-
methylamines in accordance with the reaction mechanism shown. CO, H2 and higher
organics are reaction by-products. The split of MMA, DMA and TMA in the reactor product is
close to equilibrium (if the catalyst is of the acidic macro-pore type). The equilibrium split
depends on the ratio of nitrogen to carbon in the feed. If DMA is desired in higher-than-
equilibrium quantity, then excess MMA and TMA can be recycled.

The distillation section consists of four columns. The reactor product is fed to the NH3
Column where ammonia is separated overhead as an azeotrope with TMA and returned to
the synthesis section. The bottoms stream consisting of amines and water is fed forward to
the Extraction Column where TMA is taken from the overheads and drawn as product of
recycled to synthesis. The bottoms, containing predominantly MMA, DMA and water, are fed
to the Dehydration Column. MMA/DMA is taken overhead and produced water is taken from
the bottom stream. The mixture of MMA and DMA from the Dehydration Column is fed to the
Product Column, where pure MMA and DMA are taken as products. Ammonia and amine
containing vent gases are fed to the scrubbing system so that ammonia and amine are
recovered by recycling of the scrubber purge back to distillation. These processes include
complex aqueous and organic systems and several azeotropes. They are truly multi-
component, in contrast to a binary split of key components in the presence of other more or
less volatile minor components.

3 Introduction to the problem of methylamines distillation modelling

The first step in any development is to be very clear about the reasons why this work is being
done. So why did DPT need to go back to basics for such a well established process?

• The distillation scheme had not changed significantly for many years so it was
necessary to review how to add value.
• Published data on physical properties for the various mixtures of components in the
process were neither consistent nor comprehensive.
• The original design was derived by a combination of basic theory and operational
experience but had not been accurately modelled using modern computer simulation
programmes.
• DPT needed to be able to simulate the methylamines distillation accurately in order to
reduce design margins to a minimum.
• In other processes DPT has proven that sound process modelling enables designers
to make considerable cost savings on capital and operating costs. To omit
methylamines would have been a departure from normal business practice.

It was soon realised that this was a difficult task. Methylamines distillation, particularly where
water is present is a very difficult system to model . To model properly with high accuracy
over a wide range requires a complex model with many interaction coefficients. It was found
that commercially available regression packages do not have adequate capabilities (without
excessive user programming) to rigorously model the methylamines system correctly.
However, this paper describes how commonly available thermodynamic options were used to
model plant design and optimisation with appropriate accuracy.

Publicly available information was collected and regressed. There was no data for some
systems and for others there was only azeotropic data. It became obvious that in-house
generation of a considerable amount of VLE data was required. Tables 3.1 to 3.5 show the
published data covering azeotropes in a methylamines distillation system.

Table 3.1 DMA-TMA Azeotropic data

Pressure Boiling Mol Fraction Reference
Point
Bara °C TMA DMA
1.01 3.0 0.68 0.32 Rohm and Hass
8.39 73 0.23 0.77 Co. (1949)
26.52 Non-azeotrope Babcock (1936)

Table 3.2 MMA-TMA Azeotropic data

Pressure Boiling Mol Fraction Reference
Point
Bara °C TMA MMA
1.01 -5 0.18 0.82 Andrews and
Spence (1938)
5.15 36 0.08 0.92 Rohm and Hass
15.49 75 0.06 0.94 Co. (1949)
26.52 Non-azeotrope
1.39 0 0.20 0.80 Hacker, Lucus
and Gelbin
(1964)

Table 3.3 NH3-TMA Azeotropic data

Pressure Boiling Mol Fraction Reference
Point
Bara °C NH3 TMA
20.27 50 0.914 0.086 Issoire (1961)
40.53 80 0.919 0.081 Issoire (1961)
1.01 -34 0.888 0.112 Issoire (1961)
0.932 -37 0.887 0.113 Issoire (1961)
0.750 -40 0.879 0.879 Issoire (1961)

As can be seen there are three pressure-dependant azeotropes which contribute to the
difficulty of separating the purified amines from each other and ammonia. These azeotropes
demonstrate the strong liquid phase interactions that make the liquid highly non-ideal, and
hence difficult to model.

Table 3.4: Example of Methylamines Vapour Density

Pressure Temperature Ideal Gas Law Vapour Density Measured
Density Ref. Kenner and Felsing (1939)
25 bara 125°C 23 kg/m3 28 kg/m3

The vapour density is highly non ideal which makes modelling the system more complicated.
Methylamines and ammonia are highly hydroscopic; they partly dissociate in water to form
basic solutions.
Table 3.5: Dissociation of ammonia and methylamines in water at 25°C
System Kb at 298K Reference
NH3 + H2O <-> NH4+ + OH- 1.8 x 10-5 G Solomons (1988)
(CH3)NH2 + H2O <-> (CH3)NH3+ + OH- 4.4 x 10-4 G Solomons (1988)
(CH3)2NH + H2O <-> (CH3)2NH2+ + OH- 5.2 x 10-4 G Solomons (1988)
+ - -5
(CH3)3NH + H2O <-> (CH3)3NH + OH 5.0 x 10 G Solomons (1988)

Table 3.5 shows that DMA is the most strongly dissociated of the 4 chemicals.
Figure 3.1: Graph showing Txy diagram for MMA and water.
VLE of MMA /H2O at about 14 bara
210

190

170

150
T (°C)

130

110

90

70
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Mol Frac MMA

Graph 3.1 shows that small quantities of water have a strong impact on the dew point of
mixtures of water and MMA.

3.1 Getting the basis correct

There is little advantage in having a complicated VLE model if the fundamental physical
properties are wrong. An accurate vapour pressure model for pure components is therefore
essential. The approach was to regress accurate VLE into the model and then use the model
within the known range of accuracy. This was valid providing the experimentally measured
ranges were close to expected process conditions in industrial distillation units.

Figure 3.2 Schematic Flowchart of DPT’s approach to methylamines VLE modelling

A. Use VLE from literature to build a preliminary model

B. Use model to choose experimental data that is needed to model distillations accurately.

Use preliminary C. Experimental Use preliminary model to

model to guide determination of VLE. help assess if results seem
experiments. reasonable.

D. Regress VLE into Model.

Experimental
VLE at tray E. Test Model.
composition

Poor fit

Good fit

For designs outside

Use model for design for flow ranges of known
schemes within ranges of known Experimental
accuracy check final
accuracy. VLE at tray
results with further
composition.
experiments.
3.2 VLE data is necessary for accurate modelling

Temperature, pressure and liquid composition data (TPx data) are readily measured and can
be generated by charging known quantities into a vial, heating to a known temperature and
recording the pressure. Hence, it is possible to get accurate results quickly and without
analysis. However, this simple approach would not have produced an adequate model for the
methylamines system. It was realised that, since the vapour phase was strongly non-ideal, it
was essential to know the vapour composition accurately. The only option was to design a
system that would produce temperature, pressure, liquid composition and vapour composition
data (TPxy data).

Constant pressure data has traditionally been collected by DPT and so the following data was
measured:

• Binary data which could be regressed.

• Ternary and higher order data to test the accuracy of the model.
• An accurate mass balance over the VLE unit.
• Specific tray compositions were tested on the unit.

After analysis and regression of data from literature a preliminary model was developed. To
make the job of doing the experiments as easy as possible, results from a literature survey
were used to prepare background information on the system. This included:

• Boiling temperatures of pure components at pressure.

• Position of any azeotropes known by literature.
• Model-predicted distillation curves.

Experience suggested that, even if huge quantities of data for a distillation model were
collected, the ultimate number of trays predicted would depend on just a small fraction of this
data. This led to two important conclusions. Firstly, the critical areas of a column should be
identified so that the design does not rely on just a few experimental points. Secondly, the
data requirements can be minimised in non-critical areas. The following strategy was used to
identify the critical regions in which to generate data:

• Search literature for VLE and azeotropic data.

• Regress literature data.
• Prepare preliminary model.
• Use preliminary model to simulate the distillation.
• Identify compositions that are critical to the design of each column.
• Choose first phase of binary and higher order data that will be experimentally
determined.

It was expected that this ‘first phase’ of experiments would generate all the data needed for a
satisfactory model because the process of model improvement is cyclic. In addition minor
components build up in the distillation scheme and have a significant impact on distillation
energy requirements. Furthermore, some of the literature data was of poor accuracy and this
was not surprising given the problems that were faced during experimentation.

4 Analytical Techniques and Equipment Description

When the portfolio of amines technologies were acquired, the associated analytical
procedures were outdated and so DPT developed new procedures using modern techniques.

DPT research chemists encountered and overcame a number of technical problems when
they attempted to generate the VLE data for this system. Accurate sampling was a problem
when dealing with high pressure mixtures of methylamines. Methylamines are toxic and
flammable and they have a very low odour threshold limit. The multi-component mixture at
high pressure will flash if the pressure is let down when a sample is taken.
The first attempt at sampling used a specially designed syringe for taking samples at high
pressure. A valve on the syringe allows the contents to be kept at pressure while it is taken to
the GC. The syringe was pushed through a rubber septum that was attached to the sample
point. This method suffered from failure of the septum due to chemical attach of the rubber
by amines. Amines were released into the fume cupboard which was obviously very
undesirable, hence a redesign of the sample point was required.

Figure 4.1: A sample point and schematic diagram.

The three way valve (B) was opened to purge the capillary line. The end of the capillary line
was placed into a 2ml sealed sample container that had been charged with a solvent. The
solvent was used to stabilise the sample to prevent loss of sample due to flashing at low
pressure. The solution was then injected into the GC. There is a balance to be made
between the loss of accuracy due to dilution and the loss of accuracy due to insufficient
solvent to stabilise the sample. The quantity of solvent was varied depending on the
expected composition of the sample. In most cases, ethanol was used as the solvent.

One key factor in the successfully measurement of VLE is the design of the equilibrium flash
pot. This was based on a design which had been successfully operated on another unit.
Initially good results for binary systems that had water and one amine were obtained.
However, when the system was switched to a binary system with two methylamines the
results were poor indicating a fundamental problem in the VLE separator. The liquid head
generated by the low density amine was not sufficient to provide a liquid seal so the separator
was modified to overcome this problem.

The key modification was to increase the size of the liquid collection vessel. The enlarged
vessel ensured a liquid seal was maintained. An accurate mass balance over the rig allowed
internal checking of results giving a high degree of confidence that the results were accurate.
Product flows, composition and feed flows were measured at near-adiabatic conditions by
balancing heat supplied to the VLE separator with heat loss, by wrapping it with heating coils
and insulating it. In calibration runs, the unit gave accurate results for a known system with a
similar boiling range to the methylamines.

Like most of the larger pilot plants at DPT’s Technology Centre, the apparatus was DCS
controlled. The key benefit of this is accurate temperature control within the VLE separator.
In addition the feed pump, safety trips and level control could be left in automatic for hours so
that it was very nearly at equilibrium before samples of vapour and liquid were taken. Several
samples could be taken at a single condition.
Figure 4.1: Photo of the DPT High Pressure VLE Rig Figure 4.3 Diagram of the VLE Apparatus.

5 Confirmation of Simulation Model

The preliminary model used the ideal gas equation for the vapour and NRTL liquid activity
coefficients for the liquid. This was appropriate given the accuracy of the information
available in literature. Two different models were then tested, namely: NRTL + ideal gas
(NRTL-Ideal) and NRTL + SRK-Simsci modified (NRTL-SRKM).

The NRTL-Ideal model did not accurately predict all the binary VLE systems accurately.
NRTL-SRKM was able to get very accurate fits for all systems. The NRTL-Ideal model gave
a better fit to the higher order VLE (3 and more components), however the difference between
the fit obtained using the two models was fairly marginal. The strongly non-ideal vapour is not
modelled accurately by the ideal gas equation of state.

Confidence in the model was gained by measuring three component data. These data were
then used to test the binary fitted model and, where necessary the binary parameters were
adjusted to achieve a better fit. Experimental multi-component VLE data at very close to
actual tray compositions at several locations within the distillation columns were collected.
The table below gives an example of the closeness of fit achieve between model and
experiment on two trays in the extraction column.

Table 5.1: Comparison between experimental results and predicted tray compositions
Column Temp °C Difference between model and Exp. Compositions (mol fraction)
MMA DMA Water Ammonia TMA
Extraction 86 0.013 0.008 0.010 0.000 0.015
Extraction 92 0.006 0.016 0.011 0.000 0.002

The model results showed a good correlation with industrial plant data.

Through the exercise of modelling experimental trays it was possible to demonstrate the
strengths and weaknesses of the model. Perfect correlations between experiment plus model
and plant data were not achieved but this is a useful tool for design and for trouble shooting
or plant de-bottlenecking. This has confirmed that the approach to model selection is valid.

The VLE unit is a valuable resource for generation of multi-component high pressure VLE of
toxic substances. It has been successfully used to develop an optimised design for an
alternative process for the production of improved quality tetrahydrofuran. Moreover, the test
unit can be made available for contract design and optimisation studies of a wide range of
distillation systems.
5.1 Example Reduction in Steam Requirement for the Methylamines Flowsheet

Conventional industrial scale distillation columns for amines are relatively large compared to
their production capacity and utility costs are a significant component of the manufacturing
cost. Steam consumption also depends heavily on the split of mono, di and tri drawn as
product. The new VLE model provided confidence to reduce steam costs by $50 to $70 per
tonne by making small changes to the process design.

The original four column design by Leonard has been repeated many times but other
arrangements of distillation columns have been proposed in patents. These have been
evaluated as alternative arrangements.

US Patent 2,848,386 describes the following flow-scheme for producing anhydrous MMA from
a mixture of methylamines. If an operator requires only MMA product, then DMA and TMA
can be recycled without the need to isolate DMA and TMA. When this flow-scheme is
compared with the Leonard scheme it can be seen that there is one less column. However,
simulations using the DPT model have shown that it is difficult to get high purity MMA using
this scheme, and steam usage doesn’t represent a significant saving over the Leonard
Scheme. Having said this, there are certain combinations of reactor product splits and plant
capacity where the benefits of this arrangement are accentuated.

Figure 7.1: Copy of US Patent 2,848,386 Fig.1. Figure 7.2: Diagram from US Patent 2,999,053

US Patent 2,999,053 describes a distillation arrangement where extraction water is fed to a

number of positions in a distillation column as shown in the diagram. This arrangement has
an advantage over the Leonard flow scheme because it achieves in one column what the
Leonard scheme requires two columns to undertake. However, model simulations show that
it is actually quite difficult to remove the TMA and ammonia from the MMA, DMA, water and
methanol using this arrangement. In addition, significant quantities of DMA are lost overhead.
If TMA is not required as a product, an economic evaluation of the benefits to a particular
client can be designed for a particular reactor product split and plant capacity.

5.2 Improvements in Distillation made by DPT

DPT has a number of proprietary arrangements of distillation systems to meet the various
requirements of industrial plant operators. As described above, a producer may be interested
in a methylamines unit for a dedicated purpose such as monomethylamine for an n-methyl
pyrrolidone plant. It would not be sensible to purify the trimethylamine and dimethylamine to
high purity if these are going to be recycled. Currently, DPT is investigating how best to
manage its intellectual property in this area, but already the work described in this paper has
yielded a solution for dedicated mono, di and tri methylamine systems and various
combinations of these products. Some of these arrangements have already been
commercially proven, for example a 2-column design for an MMA dedicated plant.
Other improvements resulting from an accurate simulation capability are:
• The unit can be heat confidently integrated with the rest of the plant.
• DPT can accurately predict tray temperatures and compositions.
• Plants can be effectively debottlenecked to maximise output from existing equipment.

6 Impact on Other DPT Technologies

The analytical procedures described in this paper have also been used to accurately
undertake catalyst research into a range of methylamine and higher alkylamine aminations,
enabling a greater understanding of catalyst performance over a wide range of operating
conditions. It is apparent that the accurate handling of amines in the laboratory is not easy
and requires specific techniques and these skills are currently both within the methylamines
process and through aminations to a wide range of derivatives. Since acquiring the amine
technologies DPT has been able to significantly reduce the costs of amine production and
also the environmental impact of these processes. Two notable additional derivatives to the
portfolio of licensable technologies are the solvent n methyl pyrrolidone (NMP) and methyl
diethanolamine (MDEA) which is increasingly the solvent of choice for gas treatment.

These techniques are not unique to amination processes and similar approaches have been
taken to many of the company’s licensed technologies that include pressure or vacuum
distillation. DPT has demonstrated that an accurate process simulator is a cornerstone for
developing and commercialising process technologies in a cost effective, efficient and timely
manner.

7 Conclusion

Davy Process Technology has modelled carefully selected physical phenomenon in a range
of distillation schemes by rigorously using the most the up-to-date modelling theory. Through
appropriate use of laboratory tests, a practical model has been developed that is suitable for
designing methylamine plants. This pragmatic and practical approach is a cost-effective
balance between the pursuit a ‘perfect’ model and maintaining control over research costs.
The use of such models with targeted experimental data provides an extremely high level of
confidence to optimise plant design.

8 References

Andrews, E.A., Overbrook, Spence, L. R., 1938. Separation of Methylamines. U.S. Patent
2,126,600.
Babcock, D.F., 1936. Separation of Dimethyl and Trimethylamines by Distillation. U.S. Patent
2,049,486
Graham Solomons, T.W., 1988. Organic Chemistry, Fourth Edition, John Wiley & Sons, New
York. pp. 886-887.
Hacker, Lucus and Gelbin, 1964, Chem. Tech. (Berlin) 16; C.A. 61, 1528.
Horsley, L.H. (1973) Azeotropic Data-III, Advances in Chemistry Series 116., American
Chemical Society, Washington, D.C.
Issoire,J. and Pfertzel,R., 1961. Chim.Ind. Paris, 86 p.101
Kenner, C.T. and Felsing, W.A., 1939. The Pressure-Volume-Temperature Relations for
Gaseous Monomethylamine. J. Am. Chem. Soc., 61(9), pp.2457-2459.
Rohm and Hass Co., 1949. Physical Properties of the Methylamines. 1949.
Verschueren, K., 1977. Handbook of Environmental Data on Organic Chemicals. Van
Nostrand Reinhold Company, New York.