1993 - Distillation Column Targets
1993 - Distillation Column Targets
1993 - Distillation Column Targets
00
Printedin Great Britain.All rightsreserved Copyright0 1993 Pergamon Press Ltd
(Final revision received 10 August 1992; received for publication 30 September 1992)
Abstract-This paper relates to the design of distillation columns. In particular, it relates to the optimis-
ation of a base case design. Column optimization involves options such as different reflex ratios, pressures,
side condensing/reboiling and feed preheating/cooling. We need to establish heat loads and temperature
levels for such modifications and identify the best combined options. The paper presents a methodology
based on a combination of thermodynamics and practical aspects of column modification. The method-
ology gives the engineer targets for design options ahead of design. In addition, it provides targets for
the best combination of options.
Alongside heat load and temperature targets, the methodology clarifies the effect of design modifications
on column capital cost, also ahead of design. The methodology is applicable to nonideal multicomponent
systems and complex distillation configurations. There are downstream applications involving distillation
c&nnn integration in the overall prockes and column sequencing. The-methodology has b&n tested in
industrial applications involving ethylene, LNG and refinery columns.
549
550 V. R. DHOLE and B. LINNHOFF
Side
Reboiler ; Reboiler
Condenser
w Condenser -
H H
T (C)
T (C)
t
170
I
, I I
30 40 70 k (MW)
(a) (b)
00 Stages
00 Side Exchangers
Losses due to
l Sharp separation
0 Chosen column
ABCD
- configuration
b
l Pressure drop
CD
relative volatilities or compositions which can be of column design and modification. These losses
tedious. The procedure suggested by Teranova and include inevitable feed losses, losses due to sharp
Westerberg (1989) does not require a reversible separation, chosen distillation configuration (e.g.
scheme, however, it is not clear whether the model multiple products, single column, side stripper etc.),
satisfies the sharp separation limitations. The pro- pressure drop losses, etc. The column at PNMTC will
cedure also requires iterations for relative volatilities still require infinite stages and infinite side exchangers
and a separate evaluation of minimum reflux and as shown in Fig. 5.
bubble point calculations. To bring the current
2.2. Generating the CGCC
models into more realistic domains involving variable
relative volatilities, would make them too difficult for The new approach proposed in this paper utilises
practical application. results from an already converged column simulation.
In the literature, the interactions between various Normally the outputs from simulations provide
column design modifications such as feed preheating/ molar flows and compositions on a stage-by-stage
cooling, side condensing/reboiling and reflux changes basis. Let us consider a light and heavy key model.
are not clearly discussed. Furthermore, the trade-offs By key components we mean the two main separating
between energy benefits due to these modifications components in the feed mixture whose separation is
and column capital cost are hard to assess ahead specified. The more volatile of these components is
of design. The CGCC developed by Terranova and the light key and the less volatile is the heavy key
Westerberg (1989) evaluates the effect of different (King, 1980; Kister, 1992). As a close approximation
feed vapour fractions. However it requires the iterative to PNMTC, we simultaneously solve the equilibrium
procedure for the CGCC to be repeated for different
feed conditions which can be cumbersome.
2. PROPOSED APPROACH
and the operating line equations for the key compon- Where G* and L* represent the molar flows of
ents. Figure 6 illustrates the operating line equations the equilibrium vapour and liquid streams. After
for light and heavy keys. For equilibrium compositions calculating Xo,_ and HLtin, we can set up enthalpy
we use the stage-by-stage compositions as provided balances at each of the stage temperatures and
by the simulator output (asterisked in Fig. 6). The evaluate the net enthalpy deficit (Hdcf) at each of these
compositions of liquid and vapour streams emerging temperatures (see Fig. 7a). Thus for an envelope
from the same stage are the equilibrium compositions before the feed stage we obtain:
at the stage temperature. Thus in order to solve the
ffdcf = %, - ci,,. + HD 7 (3)
equilibrium line and the operating line equations
simultaneously, all we need to do is to incorporate the and for an envelope at and after the feed stage we
equilibrium compositions of the vapour and liquid obtain:
streams emerging from the same stage into our mass
ndef = w,in - &,in + cnD - %ed). (4)
balance equations. i.e.
Figure 7b demonstrates how the individual enthalpy
G,, Y,* - Lti,,X; = D,,
deficits are cascaded to construct the CGCC. The
Gtin Y; - L,,X;: = DH . (1) values of the stage temperatures (T,, T, etc.) and the
corresponding heat deficits [H,,Al), &&) etc.] are
These equations establish the minimum vapour
flow (Gmi,,) and the liquid flow (L,,) at the stage
temperature.
In order to obtain the temperature-enthalpy
representation for PNMTC, we need to express
the minimum vapour and liquid flows in terms of
enthalpies. Usually simulation outputs also provide
stage-by-stage vapour and liquid enthalpy values. As
mentioned earlier, the vapour and the liquid streams
emerging from the same stage (asterisked in Fig. 6)
are in equilibrium with each other. The enthalpies
of these equilibrium vapour and liquid streams are
termed here as WE and W,*, respectively. The
enthalpies for the minimum vapour and liquid flows
(HGminand H,.) are obtained from ff,$ and H,t by
direct molar proportionality, i.e.
Stage
Reboiler Reboiler
Condenser Condenser
plotted in the T-H dimension as shown. The algor- values, the remaining procedure remains unchanged.
ithm used for developing the cascade is identical to An approach for an N-component model has also
the problem table algorithm introduced by Linnhoff been developed (Dhole, 1991). However, even the
and Flower (1978). The feed enthalpy strongly influ- simple model described here is useful for many
ences the shape of the CGCC near the feed stage. practical applications.
The CGCC usually shows a pinch point near the feed
stage. 2.3. Use of the CGCC
We are ready so summa&e the procedure for The CGCC is readily used for targeting for
obtaining the CGCC. We start from a converged different possible column modifications.
simulation. We first obtain the minimum vapour and Figure 9 describes the targeting procedure.
liquid flows, at each stage temperature, by solving the A horizontal distance between the CGCC pinch
two simultaneous equations (1). These minimum point and the vertical axis represents the scope
Xows are then expressed as equivalent enthalpy flows for reduction in reflux ratio (Fig. 9a). The CGCC
to obtain the net heat deficit at the stage temperature. pinch point indicates the minimum reflux condition
The heat deficits are cascaded in order to obtain the for the column. As we reduce the reflux ratio, the
CGCC. The CGCC can alternatively be plotted as CGCC will move towards the vertical axis, thus
stage number vs enthalpy (Fig. 8). This version is reducing the reboiler as well as the condenser
particularly useful for identifying the appropriate load.
feed condition. The next modification to consider is feed condi-
The procedure here has been described for a simple tioning. Inappropriate feed condition usually causes
column involving a single feed and two products. It a sharp enthalpy change in the profile near the feed
is easily extended for columns involving multiple location. For example, a feed which is excessively
feeds and products and for complex column arrange- subcooled, causes sudden quenching. This will
ments (Dhole, 1991). As regards number of compon- result in a sharp enthalpy change on the reboiler side.
ents, the description here considered a single choice Such a sharp enthalpy change is particularly easy to
of light and heavy keys. The procedure can be applied see in the stage-enthalpy representation. As shown
to different choices of key components in different in Fig. 9b, the extent of sharp enthalpy change on
sections of the column or to several components the reboiler side determines the approximate heat
grouped together as light and heavy keys. Grouped load for feed preheating. Analogous arguments apply
components are particularly useful for refinery for feed cooling. Successful feed preheating and cool-
columns. The separating light and heavy key com- ing will reduce the reboiler and condenser loads,
ponents can be identified from the stagewise compos- respectively.
&ions (obtained from simulation output). The grouped After feed conditioning, we consider side con-
light and heavy key compositions are evaluated by densing/reboiling. Figure 9c describes CGCCs which
summing the light and heavy key compositions, show potential for side condensing and reboiling.
respectively. These grouped key compositions are Feed conditioning always offers a more moderate
then used in equation (1) to obtain G,, and Lmin temperature level than side condensing/reboiling.
Distillation column targets 555
Reflux
Modification
Stage
Reboiler
Feed
Conditioning
Side
Condensing
/Reboiling
Condenser H
CGCC (Temp-H)
Targeting for side Targeting for side
-condensing -reboiling
Also, feed conditioning is external to the column appropriate feed stage location should be identified
unlike side condensing/reboiling. Thus the sequence before targeting for any column modification.
for considering different column modifications is
2.4. Column Composite Curves (CCC)
recommended as follows:
So far, we have considered column modifications
1. Reflux and pressure modifications.
mainly from an energy standpoint. Modifications
2. Feed preheating/cooling.
were aimed at reducing the excess driving forces in
3. Side condensing/reboiling.
the column. How will the modifications affect the
While setting the targets for the above-mentioned number of stages? We will now attempt to relate
modifications it is assumed that the feed stage driving forces in the column to the number of stages.
location for the column has been appropriately Figure 10a shows the driving force for a stage at
chosen beforehand. Inappropriate feed positioning temperature T2_ Vapour rises from the lower stage
could cause sharp enthalpy changes in the CGCC at temperature T, and mixes with the liquid coming
similar to feed preheating and cooling. Therefore, from the previous stage at temperature T,. The
V. R. DHOLE and B. LINNHOFF
Reboiler
1Condenser
&boiler
H Ii
Cokarm Grand Cofnposite Curve Column Composite Curves
Fig. I 1. The CGCC and CCC relate to each other as in pinch analysis. The CCC can be constructed from
the CGCC.
Distillation column targets 557
mixing results in the transfer of enthalpy from the are taken into account. Since the approach directly
vapour to the liquid stream and consequently loss of takes the equilibrium compositions from a simulator,
heat and mass transfer driving forces. The tempera- the various nonldealities are considered beforehand.
ture of the mixture is T2 which is our stage tempera- This offers an improvement over the current methods
ture. The construction is repeated for each stage. The which are mainly based on Underwood-type relative
overall construction is shown in Fig. lob. We term volatility models.
this plot the Column Composite Curves (CCC). The targeted heat loads and temperatures have been
The CCC depict the distribution of stages in different found to be about 70-80% effective. The accuracy
sections of the column. The region between the can bc further improved by considering different keys
vapour and the liquid composites represents com- in different sections of the column. In the case of a
bined heat and mass transfer loss in the column large number of components (for example, refinery
(Dhole, 1991). As we increase the number of stages, columns) or low concentrations of the key com-
the CCC will become tighter, reducing the heat and ponents, higher accuracy is obtained by grouping the
mass transfer loss in the column and vice versa when lighter and heavier components. These component
we reduce the number of stages. Thus, the CCC groups are then considered as light and heavy keys
provide a link between driving forces and the number as previously discussed in Section 2.2
of stages in the column. The approach is simple to use and is easily imple-
mented through simple software either stand-alone or
2.5. Relating CCC with CGCC within a simulator environment. More importantly,
it identifies the most promising column modification
How are the CCC related to the CGCC? The
options prior to any additional simulations. The
relationship is identical to that between pinch analysis
approach provides insights into energy as well as
Composite Curves and pinch analysis Grand Com-
capital cost implications_ The interactions between
posite Curves (Fig. 11). The horizontal distance be-
various options for modifications are clearly under-
tween the CGCC and the vertical axis is the same as
stood. The approach has been tested on ethylene,
the horizontal distance between the CCC. The area
LNG, refinery and organic process columns with
between the CCC equals the area between the CGCC
promising results. Typically 10-l 5 % power savings
and the temperature axis (shown as a dotted area)
have been achieved in the subambient processes such
and represents heat and mass transfer loss. The CCC
as ethylene and LNG (Dhole 1991). In the case of
are easily constructed from the CGCC as shown
above ambient applications such as refineries about
in Fig. 11. CGCC provides the energy targets for
10% fuel savings have been observed due to the use
column modifications while CCC enable the designer
of column targeting. These savings have resulted
to assess the effect of modifications on the number of
from identification of appropriate column modifica-
stages, i.e. on capital cost. A modification in a section
tions and improved column integration Very simply,
where the CCC show tight driving forces implies
the approach enables the designer to target for
high capital cost penalty. The combined use of
modifications accurately and easily prior to column
CGCC and CCC provide an assessment of both
and process design.
energy and capital cost implications of proposed
column modifications in the targeting mode. Together,
the CGCC and the CCC identify the most promising 3. EXAMPLE
design options ahead of design.
We will now demonstrate the principles of
column targeting using an example column both
2-6. Comparison with previous work for stand-alone modification and for process
In comparison with previous work on CGCC integration.
(Fonyo, 1974a; Kaibel, 1987; Franklin and Wilkinson,
1982; Terranova and Westerberg, 1989), the current Table I. Feed and product specifications for the example problem
approach provides a more practical, simple and reli- Top Bottom
pr0pCl-tiC.S Feed product product
able approach. We do not need to devise any revers-
Molar flow (kmol h - ) 1000.0 398.0 602.0
ible scheme for the separation (such as shown in Pressure @Pa) 2JXI.O 200.0 200.0
Fig. 4). There are no limitations on the sharpness of Temperature (C) 100.0 140.3 203.8
Vapour fraction 0.00 I.00 0.00
separation. So long as the column can be simulated, Mole fractions
we can obtain the profiles and the targets for the a-HCptanC 0.2OOOoo 0.502542 0.000@33
n-octane 0.2ooooo 0.487419 0.01oOOO
column. The approach considers an actual column n-Nonane 0.01OOw 0.325601
cor&uration/slmulation. Various inevitable losses t#-DCCaW iz%zE O.Otnw38 0.332186
PI-Cl5 0.2wooO o.cHMOOO 0.33221 I
due to the feed, pressure drop, sharp separations etc.
558 V. R. DHOLE and B. LINNHOFF
Enthalpy
-1l-l
0 9 18 27 36 45 54 63 72 81 90 0 180 360 540 720 9oc
Enthalpy Enthalpy
147C
Q side
180C
Q side= 5
203.8%
Qreb=33.2 1
3.1. SingIe coiwrvl existing column (with 18 stages) along with its CGCC
and CCC. The CGCC shows scope for improvement
Table 1 shows the feed and the product specifica- in the reflux ratio by about 20-22MMBtu h-l.
tions for the example column. Figure 12a shows the Correspondingly, the CCC show a large driving
Distillation column targets 559
force loss throughout the column. By increasing the whether for stand alone column design or for process
number of stages to 30 the reflux ratio is improved integration.
by 21 MMBtu h-l. Figure 12b shows the resulting
coiumn after reflux modification along with its 4. CONCLUSION AND SIGNIFICANCE
column profiles. The CGCC has moved closer to the
Starting from a converged simulation we are able
vertical axis due to the reflux improvement and
to develop column profiles for columns with multiple
the CCC indicate reduction in the driving force loss.
components, complex column configurations and
The CGCC shows a sharp enthalpy change on the
variable relative volatilities. The profiles identify
reboiler side (about 25 MMBtu h-l). This indicates
appropriate column modifications and essentially set
scope for feed preheating. Although the CCC have
targets in terms of heat loads and temperatures for
become tighter, the driving force distribution is not
the modifications. The designer is easily made aware
uniform. On the reboiler side, the CCC could he made
of energy cost and capital cost implications. The
tighter while on the condenser side the CCC could
combined use of column targets and pinch analysis
he made wider. Based on the column targets,
curves for the background process enhances the
we introduce feed preheating of 25 MMBtu h-.
reliability of column integration in comparison to
Figure I2c shows the simulation results after the
conventional pinch analysis.
feed preheating modification and the corresponding
column profiles. The reboiler load is reduced by
NOMENCLATURE
18 MMBtu h- indicating that the feed preheating
CGCC = Column Grand Composite Curve
modification has been effective. The CCC have
CCC = Column Composite Curves
hecome tighter. However, the driving force distri- D, = Top draw heavy key molar flow
bution has become more uniform. The use of driving D,_ = Top draw light key molar flow
G= Actual molar vapour flow
force along the column has improved. The CGCC
G,, = Minimum molar vapour flow
shows further scope for side condensing and side G* = Molar flow of equilibrium vapour stream
reboiling of 10 and 15 MMBtu h- respectively. HD = Enthalpy of top draw
Accordingly, we introduce a side reboiler and a W,, = Enthalpy deficit
H fssd= Enthalpy of feed
side condenser modification. Figure 12d shows the Hz = Enthalpy of equilibrium vapour stream
simulation results. The simulation results confirm w GE+,. =
Enthalpy of minimum vapour flow
= Enthalpy of equilibrium liquid stream
the targets.
Ht
17~~~.= Enthalpy of minimum liquid flow
L = Actual molar liquid flow
3.2. Process integration Lmin= Minimum mola; liquid flow
MTC = Minimum thermodvnamic condition
We now consider the background process.
PNMTC = Practical near minimum thermodynamic con-
Figure 13a shows the overlap of the example dition
column (as a box) and the background process. X, = Mole fraction of heavy key in liquid stream
Again, note that the CGCC has been plotted in X+, = Equilibrium mole fraction of heavy key in
liquid stream
the reverse enthalpy direction (from right to left). XL = Mole fraction of light key in liquid stream
One might reduce the column pressure to avoid Xf = Equilibrium mole fraction of light key in liquid.
stream
overlap. However, considering the CGCC, we
Y, = Mole fraction of heavy key in vapour stream
identify reflux and preheating modifications as Y$ = Equilibrium mole fraction of heavy key in
appropriate. The use of column targets makes vapour stream
Y, = Mole fraction of tight key in vapour stream
the designer fully aware of modification potential,
Yt = Equilibrium mole fraction of light key in vapour
stream
REFERENCES
Fonyo Z., Thermodynamic analysis of rectification. II. King C. J., Separution Processes. McGraw-Hill, New York
Finite cascade models. Inf. Chem. Engng 14, 203-210 (1980).
(1974b). Kister H. Z., Distillation-Design. McGraw-Hill, New York
Franklin N. L. and M. B. Wilkinson, Reversibility in the (1992).
separation of multicomponent mixtures. Trans IChemE Linnhoff B. and J. R. Flower, Synthesis of heat exchanger
60, 276-282 (1982). networks I. Systematic generation of energy optimal
Ho F. G. and G. E. Keller, Process Integration. Recent networks. A/C!& Jl 24, 633642 (1978).
Developments in Chemical Process and Plant Design Linnhoff B., H. Dunford and R. Smith, Heat integration of
(Y. A. Liu, H. A. McGee and W. G. Epperly, Eds). Wiley, distillation columns into overall processes. Chem. Engng
New York (1987). Sci. 38, 1175-l 188 (1983).
Ishida M. and T. Ohno, Application of energy-direction Naka Y., M. Terashita, S. Hayashiguchi and T. Takamatsu,
factor diagram for analysis of a distillaticn column. An intermediate heating and cooling method for a
J. Chem. Engng. Japan 16, 281-287 (1983). distillation column. J. Chem. Engng Japan, 13, 123-129
Kaibel G., Ph.D. Thesis, Technical University Munich (1980).
(1987). Terranova 8. E. and A. W. Westerberg, Temperatureheat
Kaibel G., Distiilation column arrangements with low diagrams for complex columns. 1. Intercooled/interheated
energy consumption. IChemE Symp. Ser. 109, 43-59 distillation columns. Inr. Engng Chem. Res. 28, 1374-1379
(1988). (1989).