DESIGN pitfalls
Clayton E.
Jones, Nathan
A. Hatcher and
Ralph H. Weiland,
Optimized Gas
Treating, Inc.,
USA, examine an
operational pitfall
involving a
split-flow absorber
in an LNG unit.
o prepare gas for liquefaction, its CO2 content is
typically reduced to 50 ppmv or below by absorption
into an amine treating solvent. Most solvents are
based on N-methyldiethanolamine (MDEA) promoted with
lesser amounts of piperazine, although 2-(2-aminoethoxy)
ethanol, known commercially as Huntsmans
DIGLYCOLAMINE agent (DGA) and BASFs ADEG have
found use in this application.
There are several pitfalls in designing the CO2 removal
system:
Absorbers that use fast reacting solvents such as
piperazine-promoted MDEA may be subject to instabilities
should operating conditions deviate significantly from
process licensor recommendations.
Sometimes energy consumption can be greatly reduced
and solvent usage optimised if the absorber is operated
with fully-lean solvent flowing at a low rate to the column
top, and a larger semi-lean solvent flow to a location part
way down the tower. This is a split-flow arrangement
that may allow the top part of the absorber to be made
Reprinted from APRIL 2014
LNGINDUSTRY
considerably smaller in diameter than the bottom part,
thereby saving shell and internals costs. However, there is
a possible effect on mass transfer performance.
Split-flow plants can be especially sensitive to departures
from normal operating conditions, to the extent that
even a slight change in a crucial process parameter can
lead to failure-to-treat by a very wide margin. In these
cases, small changes do not lead to the expectedly small
responses in performance.
These pitfalls are by no means unique to treating gas in
LNG production. For example, ammonia production where
CO2 is removed to a few hundreds of ppm is subject to the
same concerns. Hydrogen production and the manufacture of
various synthesis gases are other examples. The
commonality, however, is the removal of CO2 to
concentrations measured in tens and hundreds of ppm using
solvents having fast reaction kinetics with CO2.
The first item in this list, column instabilities, has been
addressed for both absorbers and regenerators in other
articles.1, 2 The focus of this article is on an operational pitfall
involving a split-flow absorber in an LNG unit. First, some of
the basic principles on which mass transfer rate-based
simulation relies are explained. Then, a detailed case study is
used to expose a design and operational issue that caused an
otherwise excellent design to fail.
Figure 1. Schematic of mass transfer between vapour and
liquid being contacted within the biphase of an absorber.
LNGINDUSTRY Reprinted from APRIL
2014
Mass transfer rate-based
simulation of amine units
There are two fundamentally different methodologies
for simulating any column used to separate mixtures of
chemical species: ideal-stage and mass transfer rate-based.
Calculations employing ideal stages have been used for
about 80 years. The method is so old and well understood
that it needs no exposition here. However, it suffers from
a serious drawback to connect ideal stages to real trays
and tower packing, numerical values for tray efficiency and
height equivalent to a theoretical plate (HETP) are required.
Unfortunately, in amine treating, there is no reliable way to
determine the right values, other than using good quality data
from an already operating plant. Of course, for a grassroots
design there simply are no data. Thus, while the ideal stage
approach can be used to fit the ideal-stage model to data,
it must be viewed strictly as a regression or fitting model,
one that completely lacks any predictive power. Inability
to reliably predict performance leads to uncertainty and
oftentimes gross overdesign and this is a serious weakness.
One of the easiest ways to understand the principles of
mass transfer rate-based simulation is by analogy with heat
transfer. Heat exchangers have been designed this way for so
long that after a first course in heat transfer, little thought is
given to the fact that the methodologys basis is a heat
transfer rate model. The simplest case is a shell and tube
exchanger heating water with condensing steam on the shell
side.
The shell side temperature is constant at the saturation
temperature of the steam, but on the tube side the water
temperature gradually rises along the length of the tubes. If
this heat exchanger were modelled as a single ideal stage,
the outlet water temperature would be in equilibrium with the
steam so it would be at the steam temperature regardless of
the physical size of the exchanger itself. The exchangers
efficiency in this model would be the value needed to achieve
a match with the real measured outlet temperature. The
idealised model had to be corrected empirically, after the fact.
This certainly cannot be called predictive it is just a fit of a
poor model to measured data. Could this model be used to
design another exchanger? Yes, but only if the conditions
were the same. Could it be used to predict performance at
other operating conditions (different water flow rate, different
steam temperature)? Possibly, but with decreasing reliability
as conditions move away from those of the original, matching
case. This is patently not the best way to design heat
exchangers. Yet, it is still the most commonly used method
for designing much more expensive and complex mass
transfer equipment.
Modern heat exchanger design is done using
sophisticated software. However, the underlying principles
have remained largely unchanged since long before the
publication of Kerns classic work nearly 65 years ago.3
Whatever the method used for carrying out the calculations,
be it hand calculation or computer programs, heat exchangers
are designed using heat transfer rate models. Correlations are
used that relate tube and shell side film coefficients to flows
and fluid properties. Then the heat transfer rate is calculated
using these coefficients along with temperature difference
driving forces and heat transfer areas. The engineer provides
the exchanger geometry and specifies the design conditions
(fluid flow rates and properties), and these are sufficient for
the design to be done completely without engineer-supplied
guesses. The engineer is never asked for an efficiency of any
kind, or for artificial parameters needed to patch up model
inadequacies. This is a 100% predictive model but only in
the last 25 years has mass transfer equipment design and
analysis been put on the same solid basis. Ideal-stage
calculations still dominate, despite the availability of vastly
superior methods. The penalty is often over design to
compensate for the inherent uncertainty, and sometimes
outright failure.
At its core, a mass transfer rate model focuses on the rate
of exchange of chemical components between two phases as
driven by the extent to which temperatures and species
concentrations between phases are not at equilibrium. This is
the very antithesis of the ideal-stage model with its
assumption of perfect equilibrium, not disequilibrium. Figure 1
shows schematically two phases being contacted in some
portion of the biphase on a tray or in a segment of packing
within a column. The main premise is that the mass transfer
process is characterised by interfacial areas and mass transfer
coefficients in the vapour and liquid phases on either side of
the interface. In no sense are these parameters adjustable any
more than are the film coefficients used in heat transfer. The
flow of high and low viscosity fluids across trays and over
packing agitates the fluids and results in specific values for
mass transfer coefficients.
Engineers are not free to pick and choose coefficient
values to obtain the preferred result or some preconceived
notion of what the separation should be. On the contrary,
values of the coefficients are a function of fluid flows, fluid
properties, and the mechanical design of the specific tower
internals as expressed in well-established correlations. The
parallel with heat transfer is extremely close. Tower
performance is predetermined and predictable in exactly the
same way and to exactly the same extent as heat exchanger
performance. Every tray in an operating plant is a real tray.
The continuous nature of contact in a packed column is
handled by dividing the height into a large number of short
segments and numerically integrating across the column.
Each short segment has its own unique characteristic mass
transfer parameters that depend on the particular packing
(vendor, brand, material, size) in each segment and on the
fluids and their flow rates. Both the simulated and real
separation a column achieves depends directly on what
specific internals are in the column. Mass transfer rate-based
simulation has the authenticity that simply cannot be
achieved with theoretical stages regardless of
embellishments. This is the basis for the ProTreat simulator,
the original mass transfer rate-based tool used for the
simulation of a wide variety of gas treating processes.
Case study: split flow plant
with limited solvent cooling
Case studies are one of the most effective ways of
understanding some of the pitfalls in design. This particular
situation involved the CO2 removal section of an LNG plant
using piperazine-promoted MDEA to remove CO2 from
predominantly methane to below 50 ppmv. Focus is on the
absorption side of the process. Although it could have been
built as a single tower with a mid-tower solvent flow of
semi-lean (partially stripped) amine, the absorption section
was actually built as two towers in series, as shown in Figure 2.
Rich solvent flows through a cross-exchanger to the
regenerator where it is contacted with steam stripping vapour
generated in the regenerators reboiler. For simplicity, the
regeneration section is not shown. The largest part of the
solvent is withdrawn after flowing only part way down the
regenerator. It is used in the absorber labelled bulk removal.
This solvent has undergone only limited stripping so its CO2
content is still fairly high. However, really well stripped solvent
is unnecessary for bulk CO2 removal where the objective is to
remove most of the CO2 without achieving anything like the
<50 ppmv specification. The 50 ppmv specification is met by
the polishing absorber downstream where a relatively small
flow of well stripped solvent is all that is needed to remove
the remaining CO2 and meet the final CO2 specification.
The 4 m dia. polishing column in this case contained two
5.5 m beds of 25 - 5 HIFLOW rings while two 1-pass bubble
cap trays with 50 mm weirs served as wash trays to remove
entrained and vaporised amine from the final treated gas.
Wash water usage was 4055 kg/hr at a temperature of 50C.
Figure 2. Process flow diagram of two CO2 absorbers
connected in a split-flow configuration.
Table 1. Inlet stream flows and compositions
Raw gas
Wash water Fully lean
Semi-lean
Temperature (C)
78.2
50
48
Variable
Pressure (bar(g))
45
45
45
45
Flow (kg/h)
328 000
4055
768 000
4 477 000
Saturated
100%
Composition
Water
CO2 (mol% or mol/ 17.5%
mol)
0.021 m/m 0.388 m/m
MDEA (wt%)
37
37
Piperazine (wt%)
Methane (mol%)
82.5
Reprinted from APRIL 2014
LNGINDUSTRY
The bulk removal absorber was 6 m dia. with two 5.5 m deep
beds of 50 - 5 HIFLOW rings. Table 1 shows the flows,
temperatures, pressures and compositions of the four inlet
streams. CO2 loading is in moles of CO2 per mole of total
amine (MDEA + piperazine).
The plant to which these data refer was unable to meet
the < 50 ppmv CO2 specification by several thousands of
ppm. An undersized semi-lean cooler combined with very hot
ambient air (used as the coolant) resulted in a minimum
semi-lean temperature of about 80C, some 10C higher than
the process licensors recommended maximum of 70C. A
simulation study of this plant was carried out using the
ProTreat mass transfer rate-based simulator. As the
simulation results of Figure 3 show, there is a critical
temperature of about 76.8C above which treating is
predicted to fail precipitously. At that temperature the bulk
removal absorber becomes thoroughly rich-end pinched to
the extent that it is fully saturated with dissolved CO2. The
approach to equilibrium at the rich end of the absorber
reaches virtually 100%. Any further increase in semi-lean
solvent temperature reduces the solvents capacity and the
Figure 3. Effect of semi-lean temperature on treating in a
split-flow CO2 absorber system.
excess CO2 that cannot be absorbed spills over into the
polishing column. Unfortunately, the spill-over is of such an
amount that the polishing column cannot cope with the extra
CO2, and the treated gas quickly fails to meet specifications.
The reason for the failure is presented in Figure 4. It is
evident that as the semi-lean temperature rises, the CO2
profile in the polishing column goes from a typical lean-end
pinched shape (72, 74, 76C), where most absorption
occurs close to the column base, to a typical rich-end
pinched form at 76.85 and 77C. At 76.75C, the CO2
profile is changing position in the column extremely
rapidly. In fact, there were times in the plant when the
absorption system rapidly went in and out of compliance
with the treating specification, caused by favourable
fluctuations in other process parameters such as gas
composition, flow rates, etc.
In this case, the process operators initially felt there had to
be something seriously wrong with the column internals for
the plant to be unable to produce a gas any better than
several thousands of ppmv CO2 compared with the 50 ppmv
design target. From a process standpoint, the cause turned
out to be a lot more subtle. One does not normally expect a
temperature change of a few degrees to cause a huge
performance loss. Indeed, when the semi-lean temperature is
low enough, changes of a few degrees have a vanishingly
small effect on overall performance. This is reaffirmed by the
flatness of the curve to the left of 76C in Figure 3 and by the
very small response of the CO2 profile to temperature
changes below 76C. However, there is an extremely rapid
change from lean-end to rich-end pinch conditions resulting
from a seemingly small temperature difference near the
cut-off temperature. This is revealed by the mass transfer
rate-based simulations.
Lean and semi-lean amine temperatures were already as
low as achievable in the plant, and solvent circulation pumps
were already operating at capacity so that, short of a larger
semi-lean exchanger, treating could be brought back on
specification only by reducing the gas flow (production).
Conclusion
This case demonstrates the power of mass transfer
rate-based simulation in diagnosing plant operating
problems. What was initially attributed to malfunctioning
tower internals turned out to be caused by inherent process
sensitivity to an (unachievable) operating temperature.
A simulation based on mass transfer rates can predict
unanticipated phenomena such as this, but a simulation based
on efficiency or HETP could not reveal this critical aspect of
the process. To create confident and successful designs and
to troubleshoot existing processes efficiently, it is vital to
know which parameters control performance in different parts
of the process.
References
1. Hatcher, N. A., and Weiland, R. H., Amine Regenerator
Instabilities, Hydrocarbon Engineering, October 2012
pp. 27 - 30.
Figure 4. Profiles of CO2 concentration in the polishing column
as a function of semi-lean amine temperature.
LNGINDUSTRY Reprinted from APRIL
2014
2. Weiland, R. H., and Hatcher, N. A., Foundations of
Failure, Hydrocarbon Engineering, December 2011,
pp. 57 - 39.
3. Kern, D. Q., Process Heat Transfer, McGraw-Hill Book
Co.: New York, 1950.
