8244 Ind. Eng. Chem. Res.

2004, 43, 8244-8250

Design and Control of Distillation Columns with Intermediate

Reboilers
William L. Luyben†
Process Modeling and Control Center, Department of Chemical Engineering,
Lehigh University, Bethlehem, Pennsylvania 18015

Distillation columns in which the components being separated have widely different boiling points
feature large temperature differences between the condenser and reboiler. If cooling water is
used in the condenser, the base temperature of this type of column is often quite high, requiring
the use of expensive high-pressure steam. Energy costs can sometimes be reduced by using two
reboilers. One at the base of the column uses high-pressure steam. A second at an intermediate
tray in the stripping section can use lower-pressure steam because the column temperature is
lower at that location. This paper compares the steady-state design and the dynamic control of
a conventional single-reboiler distillation column with a column having both intermediate and
base reboilers. Design involves determining the optimum heat removal in the intermediate
reboiler, which affects both the column diameter and the rates of consumption of the two types
of steam, having different costs. Control involves handling the large temperature difference in
the column by the use of “average temperature” control and developing a control scheme to
effectively use the additional control degree of freedom.

1. Introduction This means that the temperature of the heating source

in the reboiler is much higher than the temperature of
Despite the predictions of many self-proclaimed vi- the heat-removal sink in the condenser.
sionaries during the last several decades that “distilla-
tion is dead”, the unit operation of distillation remains Many distillation columns are designed to have reflux-
the premiere separation method in the chemical and drum temperatures high enough (minimum of 110 °F)
petroleum industries. Its wide application results in a so that cooling water can be used in the condenser. This
very large variety of distillation columns. Some operate temperature and the distillate composition fix the
at high pressure and others at low pressure. Some column pressure. In these easy separation columns, both
operate at high temperature and some at low temper- products are usually produced at very high purity. The
ature. Some require many trays and others few trays. bottoms is typically a final product that requires high
This variability results in many different types of purity. The distillate is a recycle stream whose purity
columns and many different control structures. Many is less critical. Because the separation is an easy one,
of these are discussed by Luyben.1 it does not cost much to make high-purity recycle, which
When the boiling point temperatures of the compo- helps to reduce recycle rates and the reactor size. This
nents to be separated are close together, the separation means that the column pressure is essentially the vapor
is difficult. The resulting column has many trays and pressure of the lighter component (in a binary system)
operates with a high reflux ratio. The propylene- at 110 °F.
propane separation is the most important industrial The pressure at the base of the column is equal to
example. The temperature difference between the reflux the pressure in the condenser plus the pressure drop
drum and the reboiler is quite small. Heat integration through the trays. The base temperature is essentially
or vapor recompression can be considered for this type the vapor-pressure temperature of the heavier compo-
of column because of the relatively small temperature nent at this pressure. Thus, the reboiler temperature
change over the column. can be quite high because the heavy component is high
When the boiling point temperatures of the compo- boiling, and a high-temperature heat source is required.
nents to be separated are far apart, the situation is
High temperature means that high-pressure steam
exactly the reverse. The separation is an easy one, so
must be used, or in cases with temperatures above about
the resulting column requires few trays and operates
500 °F, a fired reboiler or a hot oil system must be used.
with a low reflux ratio. Examples of this situation occur
in chemical plants where the effluent from a reactor is High-pressure steam is more expensive than low-
a mixture of a light component, typically one of the pressure steam. For example, Turton et al.2 give prices
reactants, and a heavy component, typically a product. for steam at several pressure levels: 87 psia of steam
This mixture must be separated so that the light is $8.20/106 Btu, 160 psia of steam is $8.67/106 Btu and
reactant can be recovered and recycled. Distillation is 609 psia of steam is $10.40/106 Btu. Thus, there is an
often used because of the easy separation. economic incentive to substitute as much low-pressure
In these easy separation systems, the temperature steam as possible for high-pressure steam.
difference between the reflux drum and the reboiler is This substitution can be achieved by designing the
very large because of the difference in boiling points. column to have two reboilers, one in the base and one
up higher in the column. The base reboiler will have to
†
Tel.: (610) 758-4256. E-mail: [email protected]. use a high-temperature heat source. However, the
10.1021/ie040178k CCC: $27.50 © 2004 American Chemical Society
Published on Web 11/20/2004
 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004 8245

2. Process Studied
The chemical system selected as a numerical example
is the binary separation of propylene (C3d) from normal
heptane (nC7). The normal boiling points of these two
components are -53.8 and +209 °F, respectively. So,
there is a large difference between the reflux-drum
temperature and the base of the column.
So that cooling water can be used in the condenser,
the reflux-drum temperature is set at 110 °F. The
specified purity of the distillate is 99.99 mol % propyl-
ene. This means that the reflux-drum pressure is 250
psia. Assuming a 30-tray column with 0.15 psi of
pressure drop per tray, the pressure in the base is 255
psia. This means that the reboiler temperature is 453
°F for a bottoms product that is 99.99 mol % n-heptane.
Figure 1 gives the flowsheet of the standard single-
reboiler process. A feed flow rate of 1000 lb‚mol/h is used
with a feed temperature of 100 °F and a feed composi-
Figure 1. Single-reboiler column. tion of 50 mol % propylene and 50 mol % n-heptane.
The feed is introduced at stage 15 (using Aspen notation
of numbering stages from the condenser down the
column). The distillate flow rate is set at 500 lb‚mol/h,
and the reflux ratio is found that gives the specified
product purities. Because the separation is an easy one,
the required reflux ratio is only 0.253. This means that
the liquid flow rates in the rectifying section of the
column are very much smaller than those below the feed
tray, which has implications in the hydraulic design and
the control of the column. The reboiler uses 620 psia of
steam (490 °F).
Figure 2 gives the flowsheet of the process with two
reboilers. The intermediate reboiler is located at stage
25 in the stripping section below the feed tray. The tray
temperature at this location is 323 °F. A liquid pumpa-
round stream is removed from the column at stage 25,
heated to 375 °F using 326 psia of steam (425 °F), and
returned to the column at stage 24. Note that the
minimum differential temperature is 50 °F at the hot
end of the heat exchanger. As discussed in the next
Figure 2. Intermediate-reboiler column.
section, the flow rate of the pumparound is a design
parameter.
intermediate reboiler in the stripping section can use a In the specification of the pumparound in Aspen Plus,
lower-temperature heat source because the tray tem- it is important to set the “valid phases” of the “destina-
peratures are lower. tion” as “liquid-vapor” because the pumparound return
from the heat exchanger is about 20% vaporized.
Figures 1 and 2 compare a standard single-reboiler
column with a column having two reboilers. The specific
3. Steady-State Design
numerical values correspond to the case discussed in
the next section. The consumption of high-pressure Both the standard flowsheet and the intermediate-
steam in the standard configuration is 15.3 × 106 Btu/ reboiler flowsheet are simulated in Aspen Plus. The
h. With the intermediate reboiler, the consumption of Chao-Seader physical property package is used, and
high-pressure steam is reduced to 5.83 × 106 Btu/h in the “petroleum wide-boiling” convergence algorithm is
the base reboiler while the consumption of low-pressure selected to solve the column equations. The “design spec/
steam is 9.53 × 106 Btu/h. The annual savings in energy vary” capability of Aspen Plus is used to find the reflux
cost is about $50 000/year. In addition, as we show later ratio needed to give a distillate purity of 99.99 mol %
in this paper, the capital cost of the column is reduced propylene. The overall component balance results in a
because the diameter is smaller. This more than com- bottoms purity of 99.99 mol % n-heptane with the
pensates for the slightly higher capital cost of using distillate set at 500 lb‚mol/h.
three heat exchangers instead of two. The net result is The “tray sizing” calculations of Aspen Plus are used
a reduction in capital investment. to determine the diameter of the column. The worst-
case location depends on the flow rate of the pumpa-
In this paper, a numerical example is used to il- round, as discussed below.
lustrate the design aspects of columns with intermediate Flowsheet conditions are shown in Figures 1 and 2
reboilers and to study their dynamic controllability. The for the two systems. Details of column designs and the
commercial simulation software from Aspen Technology economics are given in Table 1. Note that the interme-
is used for both the steady-state design (Aspen Plus) diate-reboiler process requires three heat exchangers
and the dynamic studies (Aspen Dynamics). instead of the two required in the standard configura-
8246 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004

Table 1. Design Parameters for Two Cases would expect the liquid flow rate to increase by about
standard intermediate- 1000 lb‚mol/h from the rectifying section to the stripping
single-reboiler reboiler section. However, it actually increases by about 1400
column column
lb‚mol/h. This occurs because the liquid feed enters at
reflux ratio 0.253 0.253 100 °F, while the feed tray is at 157 °F. Some vapor
flows (lb‚mol/h) D 500 500
B 500 500
coming into the feed tray is condensed to provide the
F 1000 1000 sensible heat of this temperature change. Thus, there
R 127 127 is more vapor entering the feed tray than leaving it.
pumparound 2000 Because the difference between the liquid and vapor
compositions rates in the stripping section is equal to the bottoms
(mf C3d) z 0.50 0.50
(mf C3d) xD 0.9999 0.9999 flow rate, the increase in vapor means an increase in
(mf nC7) xB 0.9999 0.9999 liquid. Both the liquid and vapor flow rates increase
pressure (psia) 250 250 quite significantly from the rectifying section to the
temperatures (°F) reflux drum 110 110 stripping section, as shown in Figure 3.
stage 25 252 323
P/A return 375 There are also large changes in the liquid and vapor
base 453 453 flow rates near the bottom of the column. This is due to
heat duty (106 Btu/h) condenser 3.39 3.38 the large changes in temperature and composition that
intermediate 0 9.53
reboiler occur in this region. Note that the maximum vapor rate
reboiler 15.3 5.83 occurs at the bottom tray (stage 31), so the column
diameter (ft) 9.06 6.03 diameter is usually set by this worst-case flow.
total stages 32 32
area (ft2) condenser 1129 1127 It is significant to compare the vapor rates for the
intermediate 1192 standard single-reboiler column with the vapor rates in
reboiler the column with the intermediate reboiler. The vapor
reboiler 3066 1166 rate at the base of the intermediate-reboiler column is
capital (106 $) column 0.620 0.402
HX 0.438 0.459 less than half that of the standard column. The limiting
energy (106 $/year) 1.39 1.34 vapor rate now occurs on the trays where the liquid
pumparound circulates through the intermediate re-
TAC (106 $/year) 1.75 1.63
boiler. The vapor flow rate increases significantly at this
Table 2. Basis of Economics location because of the heat added in the intermediate
Condenser
reboiler.
3.2. Selection of the Pumparound Flow Rate. The
heat-transfer coefficient ) 150 Btu/h‚°F‚ft2
differential temperature ) 20 °F presence of the intermediate reboiler provides an ad-
capital cost ) 1557(area)0.65 ditional degree of freedom in the design. When the
pumparound flow rate is zero, we have the standard
Base Reboiler flowsheet. In this situation, the limiting column diam-
heat-transfer coefficient ) 100 Btu/h‚°F‚ft2 eter location is at the base of the column because this
differential temperature ) 50 °F is where the vapor flow rate is the highest. As the
capital cost ) 1557(area)0.65
pumparound rate increases, the vapor rates decrease
Intermediate Reboiler in the section of the column below it. However, the liquid
heat-transfer coefficient ) 100 Btu/h‚°F‚ft2 load at stages 24 and 25 increases, and this impacts the
log-mean differential temperature required column diameter.
∆TLM ) (∆T1 - ∆T2)/ln(∆T1/∆T2) Figure 4 shows how the flow rate of the pumparound
∆T1 ) 425 - T25 affects the column diameter and the reboiler heat duties.
∆T2 ) 425 - 375
capital cost ) 1557(area)0.65 The return temperature of the pumparound from the
intermediate reboiler is set at 375 °F, which gives a 50
column vessel capital cost ) 1917D1.066L0.802 °F temperature difference between the process and 326
energy cost: 620 psia steam ) $10.37/106 Btu psia of steam (425 °F). The temperature of the stream
326 psia steam ) $9.74/106 Btu
capital cost entering the intermediate reboiler is the temperature
TAC ) + energy cost at stage 25. The steam is about 20% vaporized in the
payback period
payback period ) 3 years reboiler when the flow rate is 2000 lb‚mol/h.
The results in Figure 4 show that as the pumparound
tion. The bases for heat-exchanger sizing and capital flow increases less heat is added in the base reboiler
and operating costs are given in Table 2. and more in the intermediate reboiler. The diameter is
3.1. Column Profiles. Figure 3 gives temperature, large at low pumparound flow rates because of the high
composition, and flow-rate profiles for the two flow- vapor flow rates at the base of the column. The diameter
sheets. The solid lines are for the intermediate-reboiler is large at high pumparound flow rates because of the
column. The dashed lines are for the standard single- high liquid loads on those trays (stages 24 and 25) where
reboiler column. Notice that the vapor and liquid flow the pumparound flows through the column.
rates are very different from those expected from the The minimum column diameter occurs with a pumpa-
commonly used “equimolal overflow” assumption. This round flow rate of 2000 lb‚mol/h. This is selected as the
is due to the large temperature differences and the large optimum design condition.
difference in the molar latent heats of vaporization of 3.3. Economics. The results given in Table 1 show
the two components: 5700 and 16 000 Btu/lb‚mol for that the intermediate-reboiler process has a 6.6% lower
propylene and n-heptane, respectively, at 77 °F. total annual cost (TAC). This economic advantage is
For example, consider the feed tray, stage 15. The feed primarily the result of the smaller diameter column,
is 1000 lb‚mol/h of liquid (50 mol % propylene). We which reduces the capital cost of the column shell by
 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004 8247

Figure 3. Column profiles.

Figure 4. Effect of P/A flow rate.

35%. The capital cost of the three reboilers in the difference in cost between 620 psia of high-pressure
intermediate-reboiler column is only 5% higher than the steam and 326 psia of lower-pressure steam. The larger
standard single-reboiler process because the total re- this differential, the larger the economic incentive. The
boiler area is less (because of the larger log-mean cost figures used in this paper are those given by Turton
temperature difference in the intermediate reboiler). et al.,2 which predict a differential of only $0.63/106 Btu.
Energy cost is reduced about 4%, which is (somewhat In many chemical plants, this differential is often much
unexpectedly) not as significant as the reduction in the larger because high-pressure steam can be efficiently
column cost. Of course, this savings depends on the used to drive steam turbines.
8248 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004

4. Dynamics and Control

Both of the flowsheets are “explored” from Aspen Plus
into Aspen Dynamics as “pressure-driven” simulations
after all of the parameters required for dynamic simula-
tion are calculated. Reflux-drum and base liquid holdup
sizes are calculated to give 5-min residence times at 50%
level. Proportional level controllers are used with gains
equal to 2. The Aspen Dynamics default structure of a
constant mass flow rate of reflux is used. This is
appropriate because of the very small flow rate in the
rectifying section of the column.
The standard explicit Euler integrator in Aspen
Dynamics did not perform well in the simulations,
probably because of the large difference in volatilities
between the two components. Changing to the Gear Figure 5. Fixing appropriate pumparound variables.
algorithm solved this problem.
4.1. Average Temperature Control. There are very a time equal to 0.2 h, a 20% step increase in the setpoint
large temperature changes from tray to tray in this type of the feed flow controller is made. At a time equal to 2
of column. If only a single temperature is used, the h, the setpoint is decreased back to normal. Finally, at
process gain is very large, which means the controller a time equal to 4 h, the setpoint is reduced 20%.
gain must be small and load response is poor. A common Figure 6 shows that both columns handle these large
way to overcome the problem3 of very sharp temperature disturbances well. The two graphs on the left-hand side
profiles is to measure the temperature on several trays are for the standard single-reboiler column. The two
above and below the sharp break and average the graphs on the right-hand side are for the column with
temperature signals. the intermediate reboiler. The heat inputs for both the
The trays selected are stages 26, 28, and 30 in the base and intermediate reboilers are shown. The former
standard column with temperatures 172, 333, and 446 is manipulated by the average temperature controller.
°F, respectively, giving an average temperature of 317 The flow rate and return temperature of the pumpa-
°F. In the column with the intermediate reboiler, stages round from the intermediate reboiler are held constant.
27, 28, and 29 are used with temperatures 345, 393, and The heat duty changes because of changes in the
433 °F, respectively, giving an average temperature of temperature and composition of the liquid coming from
390 °F. These average temperatures are the setpoints stage 25.
of temperature controllers manipulating the base-re- Maximum deviations of the average temperature are
boiler heat input. A 1-min deadtime is inserted in the 35-40 °F for both systems. These large changes result
temperature loops, and relay-feedback testing is used from the very large temperature difference over the
to determine the ultimate gains and periods. Temper- column.
ature transmitter spans are 200-400 °F. Tyreus- Figure 7 gives results for feed composition distur-
Luyben tuning rules are used, giving values of KC ) 0.6 bances. At a time equal to 0.2 h, the feed composition
and τI ) 7.6 min for the conventional column and KC ) is changed from 50 to 60 mol % propylene. Then at a
0.42 and τI ) 10.6 min for the intermediate-reboiler time equal to 4 h, it is changed to 40 mol % propylene.
column. Both processes handle these disturbances well.
The average temperature is controlled in both col- Figures 6 and 7 show that the intermediate-reboiler
umns by manipulating the heat input to the reboiler in case has somewhat slower dynamics. This may be a
the base of the column. The heat input to the intermedi- controller tuning issue, despite the fact that the same
ate is not manipulated because it should always be tuning procedure was used in both cases. As discussed
maximized. Therefore, the pumparound rate is fixed at in section 4.1, the controller gain is smaller and the
its maximum (2000 lb‚mol/h). We assume that the integral time is larger for the intermediate-reboiler case.
pumparound return temperature is always 375 °F The responses of the base case indicate more aggressive
because of the constant pressure of the steam on the tuning (a smaller closed-loop damping coefficient). As
hot side of this heat exchanger. The tray temperature shown in Figure 3, the column temperature profiles are
will vary dynamically with time, so the heat-transfer different. The trays selected to calculate the average
rate is not constant. temperature are also different, and this leads to slightly
4.2. Specifying the Pumparound in Aspen Dy- different dynamics.
namics. The fixed and free variables in the column
simulation must be specified to achieve the desired
5. Conclusion
operation of the pumparound during dynamic operation.
We want to keep the pumparound flow rate constant The design and control of a distillation column utiliz-
at 2000 lb‚mol/h and the pumparound return temper- ing a partial reboiler have been compared with those of
ature constant at 375 °F. These are not the default fixed a standard column. Consumption of high-pressure heat
variables in Aspen Dynamics. As shown in Figure 5, the can be reduced, and the column diameter is also
“All Variables” window in “Forms” for the column is reduced. The economic effect is a reduction in both
opened, and the following variables are set to be energy and capital costs.
“fixed” variables: Dynamic controllability is just as good in the inter-
(1) pump(1).F mediate-reboiler column as it is in the standard column.
(2) PumpAround(1).LiqOut.T Average temperature control should be used in both
4.3. Dynamic Performance. The two processes are because of the very sharp temperature profile. The
subjected to a series of feed flow-rate disturbances. At energy input in the intermediate reboiler is not used
 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004 8249

Figure 6. Feed-rate disturbances.

Figure 7. Feed composition disturbances.

as a control variable because it should be maximized is needed in the top condenser, an intermediate pumpa-
so that the use of higher-pressure steam is minimized. round/condenser using inexpensive cooling water could
This paper has considered the use of intermediate be located on a tray down far enough in the rectifying
reboilers. Of course, a similar situation occurs when section to have temperatures greater than 110 °F. The
intermediate condensers can be used to conserve the use design and control issues for this type of system would
of high-cost cooling in the top condenser. If refrigeration be very similar to those considered in this work.
8250 Ind. Eng. Chem. Res., Vol. 43, No. 26, 2004

Nomenclature ∆TLM ) log-mean temperature differential in the interme-

diate reboiler (°F)
B ) bottoms flow rate (lb‚mol/h)
∆T1 ) temperature differential at the cold end of the
D ) distillate flow rate (lb‚mol/h)
intermediate reboiler (°F)
D ) diameter of vessel (ft)
F ) feed flow rate (lb‚mol/h) ∆T2 ) temperature differential at the hot end of the
HX ) heat exchanger intermediate reboiler (°F)
KC ) controller gain
L ) length of the vessel (ft) Literature Cited
P ) pressure (psia)
P/A ) pumparound (1) Luyben, W. L., Ed. Practical Distillation Control; Van
QC ) condenser heat removal (Btu/h) Nostrand Reinhold: New York, 1992.
QInt ) intermediate-reboiler heat input (Btu/h) (2) Turton, R.; Bailie, R. C.; Whiting, W. B.; Shaeiwitz, J. A.
QR ) base-reboiler heat input (Btu/h) Analysis, Synthesis and Design of Chemical Processes, 2nd ed.;
Prentice Hall: Englewood Cliffs, NJ, 2003.
R ) reflux flow rate (lb‚mol/h)
RR ) reflux ratio ) R/D (3) Luyben, W. L. Control of distillation columns with sharp
temperature profiles. AIChE J. 1971, 17, 713.
TR ) reflux-drum temperature (°F)
TAC ) total annual cost (106 $/year)
Received for review June 16, 2004
xD ) distillate composition (mol fraction propylene)
Revised manuscript received September 17, 2004
xB ) bottoms composition (mol fraction n-heptane) Accepted September 27, 2004
z ) feed composition (mol fraction propylene)
τI ) controller integral time (min) IE040178K