When designing distillation columns, engineers focus on
selecting the required number of trays or packing height and
column diameter to achieve target performance. As a result,
distillation column feed condition is often overlooked.
Improper feed design can hinder performance and influences unit
This article will illustrate useful distillation column feed
design optimization methodologies. A case study is demonstrated
in detail and highlights how optimum feed design enhances
distillation column performance.
Technical considerations for feed location.
Selecting optimum feed location is critical to maximizing
distillation column performance. Improper feed location of a
distillation column can downgrade column performance; the
degree of separation is decreased at the same reflux/boil up
ratio or the higher reflux/boil-up ratio is required to
maintain the degree of separation.
An ideal feed location is feeding to a section of the
distillation column where column internal liquid traffic
composition is similar to feed stream composition. In this
case, it can minimize the composition gradient between feed
stream and distillation internal fluids. In distillation column operations, it
is often seen that feed compositions are changed from original
design conditions. In the case of significant deviation,
discrepancy between column internal liquid composition and feed
stream composition can increase, and results in non-optimum
feed location. Therefore, evaluating the feed location is an
essential step for a successful distillation unit revamp or
However, it is very difficult to sample and analyze the
column internal liquid traffic composition in most of the
distillation columns that are commercially operated. Instead,
process simulation modeling has been widely utilized to predict
internal liquid composition and determine the optimum feed
location in the actual industry design. As feed optimization
through simulation modeling is convenient and does not require
additional costs for field measurements and laboratory
analysis, it seems to be a very convenient procedure on the
surface. Nevertheless, reputable simulation software itself
does not promise the reliability of simulation modeling.
Inherent gaps between actual conditions and theoretical
simulation modeling should not be ignored.1 The
appropriate simulation flow sheeting methodology is necessary
to bridge between actual conditions and the simulation model.
It is common for improper simulation modeling to give
misleading results regarding the optimum feed-point location
and cause poorer column performance than expected.
Key ratio plotting is a useful tool to
evaluate optimum feed location in simulation modeling. This key
ratio is usually expressed as the mole fraction ratio of light
keys to heavy keys in a semi-logarithmic scale chart. A benefit
in using the chart is that retrograde distillation due to
non-optimum feed location can be visually identified. Optimum
feed location can be graphically selected through the chart.
Meanwhile, it should not be overlooked that key ratio plotting
only shows binary key component behaviors. Light non-key
behaviors are not recognized in this plotting.2
Relying on key ratio plotting is dangerous in determining
optimum feed location. Reviewing non-key component composition
profiles through the column and various sensitivity analyses is
also required to ensure optimum feed location.
Arranging multiple feed locations to a single distillation
column is one of the solutions for a column in which feed
compositions are frequently varied. Actual feed location can be
switched to suit varying feed composition. Switching methods
can be arranged either by manual block valves or automatic
control valves. This design is usually feasible in a trayed
distillation column. It is difficult to arrange multiple feed
locations in most packed column cases unless an extra bed is
inserted to allow an alternate bed.
Feed temperature is one of the major factors in influencing
the overall heat balance of a distillation column system.
Increased feed enthalpy can help reduce the required energy
input from the reboiler at the same degree of
separation. Installing a feed pre-heater is a very common
process option to minimize reboiler heat duty. If the feed
preheater can be integrated with other valuable process streams
(as a heating medium), overall energy efficiency of the
distillation system can be improved further. A simple heat integration in the distillation
block is heating feed streams using the bottom product.
However, increasing feed temperature does not always improve
the overall energy efficiency of a distillation unit. Excessive
feed temperature increments can cause a significant amount of
flash of heavy key and heavy non-key components at the
distillation column feed zone. In this case, a higher amount of
reflux stream is necessary to maintain required overhead
distillate purities. This augmented reflux ratio thus requires
a higher boil-up ratio. Overall energy efficiency is eventually
aggravated.3 Therefore, carefully reviewing feed
temperature and phase is critical to minimize the distillation
units overall energy consumption. Identifying feed
condition at a reliably measured feed temperature should be
conducted to evaluate optimum feed conditions. Sensitivity
analysis is a required step as well.
Internal feed-distributor design.
The importance of the internal feed distributor cannot be
ignored in distillation column feeding optimization. The
equilibrium stage basis simulator predicts distillation column
performance based on ideal mixing between feed and column
internal traffic. This means that the feed distribution quality
is not reflected in the equilibrium stage calculation result.
Therefore, feed distribution quality should be evaluated
separately to ensure that simulated distillation column
performance predictions are met.
In real distillation column operation, feed distribution
quality influences overall performance. Non-optimized feed
fluid distribution can cause non-uniform concentration across a
distillation column cross-sectional area and result in
downgrading performance. The importance of feed distribution is
emphasized when a distillation column has multi-pass trays
and/or packing. In multi-pass trays, it is essential that
uniform internal liquid-to-vapor (L/V) ratio in each section
shall be maintained.4 Also, the feed distribution
ratio shall be matched to the tray L/V ratio as closely as
possible for best performance. Otherwise, poor feed
distribution can cause an imbalanced L/V ratio in each pass. In
an imbalanced L/V ratio, the reduction of tray efficiency can
be observed and non-uniform froth height generation can reduce
overall column capacity.
Non-optimum tray layout selection can
increase the difficulty of feed distribution. A typical example
is when the distillation columns top tray is configured
with two off-center positioned downcomers in four-pass-tray
geometry. In this geometry, an internal feed/reflux distributor
has to be fed to one center and two side inlets. With a
conventional distributor design, this almost guarantees that
the split of liquid flow was not proportional to the vapor from
each pass. Poor feed/reflux distribution to the tray inlet
panel causes an uneven L/V ratio in each active area.
If the top tray is converted to the tray with two sides and
one center downcomers, feed/reflux distribution problems can be
resolved. However, this conversion requires complete layout
changes of all trays in the column, which may not be feasible
in a revamp solution. Since it requires high modification
costs, it is not a feasible solution in most cases.
Alternatively, this feed maldistribution can be fixed by an
enhanced feed distributor design. Installing controlling
orifices can split liquid flow to match the tray pass liquid
ratio. Fig. 1 displays the feed distributor for four-pass trays
with two off-center downcomers. Since controlling orifices are
positioned inside distributor arms, these features are not
revealed in Fig. 1. A drawback of this design is creating an
additional pressure drop through the feed/reflux distributor,
thus influencing pipeline hydraulics. Therefore, it is
necessary to check whether a sufficient driving force is still
maintained at a higher feed distributor pressure-drop
| Fig. 1. Feed
distributor and four-pass top tray
with two off-center downcomers.
A packed column is more vulnerable than a trayed column with
regard to feed distribution. Like multi-pass trays, a
trough-style liquid distributor with multiple primary troughs
(parting boxes) requires uniform incoming liquid distributions
to multiple primary troughs. Otherwise, uniform liquid
distribution cannot be maintained at the liquid distributor
component irrigating the next packed bed.
Proper internal feed distributor sizing is
also necessary to achieve uniform feed distribution. High feed
fluid velocity can allow more fluid flow at the end of the
distributor. Pressure drop among the entry header, lateral arm
and discharge hole needs to be optimized. It is usually
recommended that progressive pressure drop increments through
the feed distributor provide adequate feed distribution. In
case of a packed tower trough-style distributor, high inlet
liquid velocity can cause liquid splashing, allowing liquid
flows outside the primary trough(s) of the gravity distributor.
Slightly submerged guide-tube installation with properly sized
discharge holes can help prevent liquid splashing.
External feed-line configuration.
Inappropriate external feed configurations also influence
distillation column performance. If feed flow is split and
introduced to the distillation column through multiple
locations, all multiple-branch pipes shall be symmetrical.
Otherwise, feed flows are not introduced in uniform manners and
column performance is influenced. Unless external feed-line
balancing is strongly specified in process design materials,
such as piping and instrumentation diagrams (P&IDs) and
process data sheets, this line balancing may be overlooked
during the final piping design.
External feed-line geometry is also a critical issue when
two-phase fluid is formed in the feed stream line. Certain
distillation columns, such as refinery multi-product
fractionators, are inherently designed with two-phase feed
conditions. Quite a few energy saving projects specify two-phase feed
As is well-known in the industry, a two-phase feed condition
necessitates more complex design steps than a single-phase feed
condition. Undesirable two-phase flow patterns are prone to
causing unit troubles including entrainment, flow instability,
temperature and/or pressure fluctuation, hammering and pipe or
equipment erosion.5 It is generally known that a
slug flow regime should be avoided at two-phase feed inlets. A
highly aerated slug can act like frothy surge and can cause not
only column instability but also severe hydraulic pounding and
damage.6 For a particular vacuum tower inlet, a mist
flow regime can create liquid entrainment in the flash
zone.7 At high two-phase flow velocities, most of
the liquid components are turned to mist droplets and
distributed into the vapor phase. This liquid mist is prone to
be entrained in the vacuum tower flash zone and impacts
distillate product yield and/or qualities.
Understanding two-phase flow patterns and proper piping
geometry is required to avoid unit troubles. However, it is
very difficult to identify two-phase flow patterns in an
accurate manner. There is no reliable calculation method to
predict void fractions for pressure drop and fluid residence
time yet. Due to a lack of a universal model, two-phase flow
pattern prediction is varied upon the model chosen. Graphical
and empirical methods are widely used as well. Design practices
regarding piping geometry and fitting methods are available in
The case study demonstrates how column performance is
improved by distillation column feeding optimization. The
solvent-recovery column function is to separate water and
impurities from spent hydrocarbon solvent and to supply
regenerated solvent back to the main process unit. Using
contaminated solvent as the feed stream, a reflux stream
combines with the feed line and is pre-heated prior to entering
the column at one single point located at the columns
top. Water and other impurities are stripped from the
hydrocarbon solvent, and purified solvent is produced at the
bottom of the column and recycled back to the main process
unit. The original solvent-recovery column configuration is
described in Fig. 2. As capacity of the main process unit is
expanded, the capacity of the solvent-recovery column should
also be increased to supply enough solvent to the main process
unit. However, the solvent columns maximum capacity was
lower than the required capacity.
2. Original process
An engineering firm that conducted the main process unit expansion work originally evaluated
the solvent recovery column capacity expansion. The expansion study
concluded that the existing column diameter was not large
enough to handle the required column internal traffic and
implementing a larger diameter column with other periphery
equipment modifications was suggested. Since this modification
plan required a high capital expenditure and a long shutdown
period, this modification scenario was not accepted in the
given overall project schedule and budget. A more
feasible modification scenario was desired to meet the project schedule as well as the
An alternative set of unit evaluation and
optimization studies was conducted by another party.
The evaluation started with the identification of the root
cause, which is an essential step for unit optimization.
Operating condition data as well as design documents and
materials, were gathered. A preliminary study was done based on
operating conditions. The study revealed that the maximum
design and simulated duties of the feed pre-heaters were not
matched in a reasonable manner. Simulated pre-heater duties at
measured feed temperatures were much higher than the maximum
calculated heat exchangers duties. Meanwhile, all design
material data were consistent: P&ID and equipment data
sheets showed the same heat-exchanger duties. To confirm the
feed pre-heating system design, a field survey was conducted
and compared with the design materials. Surprisingly, the field
survey disclosed that all design materials were not updated to
the real equipment configurations. There were five heat
exchangers positioned in the actual feed pre-heating circuit,
while the P&ID only showed three exchangers. Through
clarifications with the technical/operating staffs, it was
found that two more heat exchangers were added to increase feed
temperature, but design document materials were not updated.
The belief was that higher feed temperature always improves the
energy consumption in a distillation unit.
To evaluate the column performance
in detail, rigorous simulation modeling was
conducted based on operating conditions. The purpose of
modeling was to identify the bottleneck point and to construct
a base model for revamping. Obtaining pertinent
operating data is necessary for reliable simulation modeling. A
review of operating data showed that daily operating conditions
were not suitable for simulation modeling. As measured stream
volumetric flowrates were not standardized, mass-balance
closure could not be reviewed. Also, gathered temperature and
pressure data were not consistent. A dedicated test run was
required to gather reliable operating data. A set of operating
data was obtained at the rate just before the maximum operating
point in a snap-shot basis. Overall mass and component balance
closure data were compiled. Each instrument position was
checked and confirmed through a field survey.8
Before performing rigorous simulation
modeling, the thermodynamic package was reviewed in
the selected commercial process simulator. It was found that
the binary interaction parameter and alpha functions were not
available between key components in the selected thermodynamic
packages data base. In this case, most of the commercial
simulators automatically choose the ideal gas law between
components and report results. It is difficult to recognize
this assumption unless detailed thermodynamic parameters are
reviewed among components. For reliable simulation modeling,
binary interaction parameters and alpha functions that were
regressed through experiment data were applied in the selected
liquid activity coefficient model. Through various sensitivity
analyses, tray efficiency of the column was quantified.
Rigorous simulation results indicated that two-phase feed
was formed at the operating feed temperature. This feed
temperature was much higher than the original design feed
temperature. The increment of feed temperature was intended to
maximize feed preheater duty and reduce overall energy
consumption. However, undesirable two-phase feed caused an
excessive hydrocarbon solvent amount in the column overhead
system. To maintain product purity specifications of the bottom
product, the reflux rate needed to be increased. A higher
reflux rate generated more vapor/liquid traffic inside the
column and limited column capacity. To destroy this vicious
cycle, the feed temperature should be decreased to maintain the
A case study was conducted to check whether the existing
feed point was optimum. Required reflux rates were simulated
with various feed points at the same bottom solvent purity.
These results are displayed in Fig. 3. This figure shows that
adding a rectification section helps to minimize the reflux
rate at the same product purity and the feeding at Stage 3
shows the minimum reflux rate. Partial key ratio plots are
highlighted among stages 1 and 5. Fig. 4 shows key ratio plot
changes as the feed point is elevated down. Monitoring key
ratio behavior with various feed points shows that retrograde
distillation was observed at the Stage 5 feeding.
3. Reflux rate vs. feed stage at same
solvent product purity.
4. Key ratio plot.
The solvent recovery unit was modified according to the
feed-optimization study. The feed tray was relocated from the
top tray to the tray matching the third theoretical stage. The
reflux stream was no longer combined with the feed stream and
was introduced to the top tray independently. The feed
temperature was reduced to maintain liquid-phase feed
condition. New feed temperature was set at a temperature
slightly lower than the simulated bubble-point temperature. In
spite of the slightly sub-cooled feed, this temperature
maintains a stable liquid-phase feeding against a minor feed
composition variation. The internal feed distributor was
designed and installed at the new feed location. Since the feed
stream contains some fouled materials, the discharge hole size
was optimized to prevent potential plugging. It was also found
that the control valve flow was reaching the critical flow
zone. A new control valve was installed to prevent the choked
flow. Modified process schemes are depicted in Fig. 5.
5. Modified process
Table 1 summarizes the solvent columns pre- and
post-modification operating data. The maximum feed rate of the
column is increased by 35%. The metered reflux rate is
significantly improved without sacrificing bottom product
purity. The new reflux ratio is only 52% of the previous
amount. The reboiler heating medium (steam)
consumption per feed is reduced by 12%. HP
1 Lee, S. H., et al, Optimizing crude unit
designs, Petroleum Technology Quarterly, 2Q,
2 Kister, H. Z., Distillation design,
McGraw-Hill Company, 1992.
3 Lee, S. H., et al., Minimizing energy
consumption in distillation units AIChE Spring national
meeting, April 2009.
4 Bolles, W. L., Multipass flow distribution
and Mass Transfer Efficiency for Distillation Plates,
AIChE Journal, Vol. 22, No. 1, January 1976.
5 Daniels, L., Dealing with two-phase
flows, Chemical Engineering, June 1995.
6 Kister, H. Z., Distillation operation,
McGraw-Hill Company, 1990.
7 DeGance, A. E., et al, Chemical Engineering
Aspects of Two-Phase Flow, Chemical Engineering,
8 Kister, H., et al., Sensitivity analysis is
key to successful DC5 simulation, Hydrocarbon
Processing, October 1998.
Lee is the manager of refining application for GTC
Technology US LLC in Irving,
Texas, and specializes in process design, simulation
modeling, distillation equipment design
and field troubleshooting for refining and aromatic
applications. Mr. Lee holds a BS degree in chemical
engineering from Sogang University, Korea.
Binkley is manager of product development for
the GTC Process Equipment Technology (PET) group in
Irving, Texas. He is a registered professional engineer
in Texas with 42 years of experience in mass transfer and
separations equipment development and applications. Mr.
Binkley has invented several separations equipment
advancement-related patents, as well as numerous product
trademarks. He earned a BS degree in chemical engineering
from Texas Tech University.