Design and Simulation of Continuous Distillation Columns

GBH Enterprises, Ltd.

Process Engineering Guide:
GBHE-PEG-MAS-605

Design and Simulation of Continuous
Distillation Columns
Information contained in this publication or as otherwise supplied to Users is
believed to be accurate and correct at time of going to press, and is given in
good faith, but it is for the User to satisfy itself of the suitability of the information
for its own particular purpose. GBHE gives no warranty as to the fitness of this
information for any particular purpose and any implied warranty or condition
(statutory or otherwise) is excluded except to the extent that exclusion is
prevented by law. GBHE accepts no liability resulting from reliance on this
information. Freedom under Patent, Copyright and Designs cannot be assumed.

Refinery Process Stream Purification Refinery Process Catalysts Troubleshooting Refinery Process Catalyst Start-Up / Shutdown
Activation Reduction In-situ Ex-situ Sulfiding Specializing in Refinery Process Catalyst Performance Evaluation Heat & Mass
Balance Analysis Catalyst Remaining Life Determination Catalyst Deactivation Assessment Catalyst Performance
Characterization Refining & Gas Processing & Petrochemical Industries Catalysts / Process Technology - Hydrogen Catalysts /
Process Technology – Ammonia Catalyst Process Technology - Methanol Catalysts / process Technology – Petrochemicals
Specializing in the Development & Commercialization of New Technology in the Refining & Petrochemical Industries
Web Site: www.GBHEnterprises.com

Process Engineering Guide:

Design and Simulation of
Continuous Distillation
Columns

CONTENTS

SECTION

0

INTRODUCTION/PURPOSE

3

1

SCOPE

3

2

FIELD OF APPLICATION

3

3

DEFINITIONS

3

4

FRACTIONAL DISTILLATION

3

5

ROUGH METHOD OF COLUMN DESIGN

4

5.1
5.2

Sharp Separations
Sloppy Separations

5
5

6

DETAIL DESIGN USING THE CHEMCAD DISTILLATION
PROGRAM

6

6.1
6.2

Sharp Separations
Sloppy Separations

6
8

7

COMPLEX COLUMNS

8

7.1
7.2

Multiple Feeds
Sidestream Take-Offs

8
8


8

DESIGN USING A LABORATORY COLUMN
SIMULATION

9

9

DESIGN USING ACTUAL PLANT DATA

9

9.1

Uprating or Debottlenecking Exercises

10

10

REFERENCES

10

APPENDICES

A

WORKED EXAMPLE

11

B

SLOPPY SEPARATIONS

26

C

SIMULATION USING PLANT DATA : CASE HISTORIES

32

TABLES

1

801E STREAM ANALYSES

26

2

MASS BALANCE

27


FIGURES

1

2

OUTLINE OF A CONTINUOUS FRACTIONAL
DISTILLATION COLUMN

4

CHEMCAD SIMULATION

14

DOCUMENTS REFERRED TO IN THIS PROCESS
ENGINEERING GUIDE

35


0

INTRODUCTION/PURPOSE

Distillation is the most developed and most applied separation process (Ref. 1).
Recent advances in computational techniques and software allow rapid
assessment of distillation requirements and as such are to be welcomed.
However before attempting a detailed computed design process engineers
should acquire a feel for the problem. This is readily achieved by applying simple
physico-chemical concepts and shortcut methods - such an approach highlights
any gross errors in the subsequent computed design.
One means of effecting a rough column design is presented and consideration is
then given to the detail design using commercially available computer programs.
Practical aspects of column design are appraised, some of the points raised
appear facile but all have caused problems in the recent past. A worked example
allows a comparison between the rough design and the computed design.
Attention is then given to actual plant columns - how data from existing columns
can be used to aid uprating or debottlenecking exercises or the design of a new
column.

1

SCOPE

This Guide covers the process design and simulation of continuous distillation
columns. It covers the calculation of the number of theoretical plates and reflux
ratio.
It does not cover the selection of column internals nor their design.

2

FIELD OF APPLICATION

This Guide applies to the process engineering community in GBH Enterprises
worldwide.

3

DEFINITIONS

For the purposes of this Guide no specific definitions apply.


4

FRACTIONAL DISTILLATION

Distillation simply involves the separation of the components of a liquid mixture
by partial vaporization of the mixture and separate recovery of vapor and residue.
A simple fractional distillation column is outlined in Figure 1. The column may be
trayed or packed. The column consists of two sections, the rectification (or
refining) section above the feed, the stripping section below the feed. Heat is
supplied to the reboiler at the bottom of the column and removed in the
condenser at the top of the column. The feed enters at some intermediate plate
in the column.
Distillate is removed from the top of the column, the purity being determined by
the number of trays in the column and the quantity of condensate returned to the
column, i.e. the reflux ratio, R = L/D.


FIGURE 1 OUTLINE OF A CONTINUOUS FRACTIONAL DISTILLATION
COLUMN


5

ROUGH METHOD OF COLUMN DESIGN

The rough method of design presented is simply a 'back of the envelope'
calculation approach. No such method is accurate for non-ideal systems and the
Fenske-Underwood-Gilliland method is recommended as one of the simplest.
This allows plateage and reflux to be readily estimated from the process
requirements and relative volatility.
As a rough guide, relative volatility may be estimated from the difference in
boiling points:
Boiling Point Difference
°C
2
5
10
20
30
50
100

Approximate Relative
Volatility
1.05
1.11
1.25
1.6
2.0
3.1
8.7

Having determined the relative volatility the Fenske-Underwood equations can be
used to calculate the minimum number of theoretical trays (Nmin) and the
minimum reflux ratio (Rmin) to effect the required separation. The Fenske
equation can also be used to determine the feed tray position, see 6.1.1. These
equations assume relative volatility does not change with composition and
constant molal overflow. Heats of mixing and heat losses are also neglected.
Details of the equations and the calculation procedure are given in GBHE-PEGMAS-603 and in Appendix A.
Although a computer implementation of the Fenske-Underwood-Gilliland
procedure exists, hand calculation is strongly recommended. The result
will form a useful check on the rigorous computer calculations - differences in the
L/V ratio should always be carefully investigated, even with non-ideal systems.
Make sure that the problem has been correctly specified and that the VLE model
is giving sensible relative volatilities. Beware of extrapolations of pure component
vapor pressure data and component interaction parameters, see Ref. 2 and 4.


5.1

Sharp Separations

For multicomponent systems with sharp separations (these generally involve
high purity in at least one of the product streams), the shortcut calculation can be
done on a binary basis. This involves establishing the two key components, i.e.
the two components which it is desired to separate. These are designated the
light key component (LK) and the heavy key component (HK). All material more
volatile than LK is treated as this component and all material less volatile than
HK is considered to be HK. The calculations must be on a molar basis.
Again a hand calculation is preferred.

5.2

Sloppy Separations

Some multicomponent separations will be of the type known as 'sloppy'. These
are less common than sharp separations and are characterized by having a
number of components in the feed which appear in both distillate and bottoms these are termed 'distributed' components. While for sharp separations the key
components will almost always be adjacent in order of volatility, for sloppy
separations there will often be several components of intermediate volatility
between the keys. Indeed it is often not obvious which components should be
chosen as keys. Sometimes the product specifications make the choice obvious.
At other times the light key is chosen as the lightest component which has a
significant presence in the bottoms and the heavy key is chosen as the heaviest
component which has a significant presence in the distillate. An archetypal
example of a column mass balance for a sloppy separation is given in Appendix
B. A shortcut calculation using PEW is a useful way of getting an initial estimate
of the component mass balance, but the estimates of Nmin and Rmin are unreliable
- they can vary significantly depending on which components are selected as
keys.


6

DETAIL DESIGN USING THE CHEMCAD DISTILLATION PROGRAM

When the number of trays, feed location and reflux ratio to effect a required
separation have been roughly established use can be made of the rigorous
distillation computer programs commercially available. These programs (Ref. 3)
will carry out a full heat and mass balance simulation including composition,
temperature and flowrate profiles. The most widely used of these is the
CHEMCAD distillation program. This program requires the following input data:
(a)

Thermodynamic data - vapor-liquid equilibria data and vapor and liquid
enthalpies.

(b)

Hardware data - the number of theoretical stages, and the positions of
feeds, products (including sidestreams) and heat exchanger (reboiler,
condenser).

(c)

Process data - feed rates and compositions, column pressures plus
product rates and heat exchanger duties to define the operation of the
column.

CHEMCAD is essentially a performance model in that there is no provision to
vary hardware data. However, a design facility is available which allows process
data (e.g. reflux ratio) to be varied to meet performance specifications (e.g.
distillate composition).

6.1 Sharp Separations
For sharp separations and ideal behavior the shortcut method will give
reasonable starting values of Nmin and Rmin .
For non-ideal systems, the use of the results from a shortcut calculation as a
starting point for the detailed design can produce a non-optimal design which
may be too close to minimum reflux. This arises because the Underwood
equation tends to underestimate the minimum reflux for non-ideal systems.
In such cases it is better to use the rigorous program to calculate Nmin by using a
large reflux ratio (approximately 100,000). For this use about twice the Fenske
estimate of Nmin and vary the distillate rate until the distillate composition meets
the design specification. Provided the calculated bottoms composition exceeds
specification, the true Nmin is then obtained by counting stages down the column
composition profile from the top until the bottoms specification has been met. An

optimal design will be near the point where the number of stages is twice the
minimum value. CHEMCAD's design facility is particularly useful in this
design process..

6.1.1 Feed Tray Position
For a binary system with a liquid feed at its bubble point, the feed should be to
the tray which has the same liquid composition. For a multicomponent feed at its
bubble point, a first estimate of the feed plate is that plate on which the ratio of
compositions of the key components in the liquid is the same as their ratio in the
feed.
For an initial estimate, these criteria may be used whatever the feed condition.
Large errors are likely only with very cold liquid or very hot vapor feeds.
Alternatively, the Fenske equation may be used to find the feed plate at total
reflux by substituting the feed composition for the bottoms composition - the
equation then predicts the number of trays above the feed. Since this is at total
reflux it needs to be multiplied by the ratio of the number of theoretical trays in
the column divided by the minimum number of trays.
None of these methods is likely to be accurate, because either the feed is not
liquid at its boiling point or the method is not reliable. It is therefore important to
search for the optimum feed position by calculating the reflux ratio required as a
function of feed plate position. The reflux ratio can be a surprisingly strong
function of feed plate position in some cases and significant savings in reboil
heat are to be had by finding the optimum.

6.1.2 Mass Balance
In a distillation with no reaction what goes in must come out, either as distillate or
a bottoms stream in a simple column.
If a feed contains 20% of the more volatile component and it is required to
recover this component as a pure distillate this cannot be achieved if more than
20% of the feed is removed as distillate.
It is important to draw up a tabular mass balance before using CHEMCAD.
Failure to do so can result in an infeasible specification with severe convergence
difficulties.

6.1.3 Heat Balance
A simple heat balance around the system, to check the computed value, can be
carried out with a knowledge of the latent heat of vaporization.
Assume 1.9 x 10 6 kJ/te for water and 0.38 x 10 6 kJ/te for other components.
In systems with large flows sensible heat requirements must be taken into
account. Again for checking purposes assume:
Specific heat capacity for organics ~0.5 cal/g °C (2.1 kJ/kg °C)
Specific heat capacity for water ~1.0 cal/g °C (4.2 kJ/kg °C)
Thus if a hydrocarbon feed at 200 te/h enters a column at its bubble point of 90
°C and 180!te/h leaves as bottoms at 160 °C, the sensible heat requirement is:
180,000 x (160-90) x 2.1 = 26.46 x 10 6 kJ
1 MW = 3.6 x 10 6 kJ
i.e. a sensible heat requirement of 7.35 MW.
Note:
In the above calculation of the sensible heat requirement no account is taken of
the low distillate rate leaving the system. This would not be the case if the
distillate rate was high relative to the bottoms rate.
Such large sensible heat requirements are often evident in extractive distillation
systems (often coupled with a liquid-liquid extraction system).
Typical input and output from the CHEMCAD distillation program for a sharp
separation are given in the worked example in Appendix A.
The output data should be examined carefully:
- Are the required tops and bottoms specifications being met ?
- Does the composition profile show a pinch condition ?
Is this the result of too many trays or insufficient reflux ?
- Is the computed heat input of the same order as that calculated using the
shortcut method!? If not check your units and/or the feed tray position.

6.2

Sloppy Separations

The characteristics of sloppy separations are described in 5.2, with an example
in Appendix B, Table 1. For sloppy separations the product specifications are
often expressed with respect to a group of components, for example the sum of
concentrations or the combined recovery.
Appendix B further outlines an example with a 7 component system -this is a cut
down version of the 28 component system in Table 1. The specifications, both
referring to the bottoms stream are:
(a)

overall recovery of components 4 & 5 to be 90%, and

(b)

total concentration of components 1, 2 & 3 to be 2 wt% max.

CHEMCAD is first used to find the minimum number of stages with a 'DESIGN'
run at a reflux ratio of 100000. The distillate rate is varied to meet the
specification 90% recovery of components 4 & 5 in the bottoms. Examination of
the liquid composition profile, Appendix B, shows that the other specification (a
total of 2 wt % of components 1, 2 & 3 in the bottoms) is met between 5 and 6
stages - condenser, reboiler and 3-4 stages in the column.
A second CHEMCAD run is then done to find the required reflux ratio with
roughly twice the minimum number of stages in the column - a total of 8 including
condenser and reboiler. Appendix B shows that both specifications have been
met. Further runs are required to find the optimum feed position, and possibly
also to explore the relationship between reflux ratio and number of stages.

7

COMPLEX COLUMNS

Sub clause 6.1 and 6.2 cover simple columns i.e. single feed, tops and bottoms
take-off. More complex columns are in common usage.
7.1

Multiple Feeds

The design is the same as covered above. The same approach can be adopted
to locate the feed points as with a single feed, taking each feed in turn. However
searching for the optimum with multiple runs will be time consuming. An
alternative approach is to plot the temperature profile for each run. When a feed
is not at its correct position, there will be a sharp change in the gradient of the
temperature profile at its feed point.

7.2

Sidestream Take-Offs

Providing that a high purity is not required an intermediate boiling component can
be recovered using a sidestream take-off. If the sidestream is taken above the
feed, normally as a liquid, the main impurities will be more volatile components.
Taking the sidestream below the feed, normally as a vapor, will result in the main
impurities being less volatile components.
The optimal sidestream tray location is most easily ascertained by means of a
few computed runs.
The purity of the intermediate product can be improved by using a sidestream
stripper (for a liquid sidestream) or rectifier (for a vapor sidestream) - see GBHEPEG-MAS-604.
Note:
The presence of ppm of an intermediate boiling component in the feed to a
column may give rise to a composition profile in which this component is present
at much higher concentrations (50% is not unknown) at some point in the
column. If necessary, consideration can be given to removing this component via
a sidestream take-off.

8

DESIGN USING A LABORATORY COLUMN SIMULATION

The measurement of vapor-liquid equilibria data for a complex non-ideal mixture
is very time consuming. An alternative approach is to simulate the distillation in
the laboratory. The column requirements can be initially determined using a
rough method of design.
The laboratory column may consist of glass sieve plates (e.g. a 2" diameter
Oldershaw column) or a column containing a suitable packing (e.g. Sulzer
laboratory scale structured packing) see GBHE-PEG-MAS-602.
The Oldershaw column is usually assumed to have a plate efficiency of 50%.
(Work in the late sixties showed that surface tension positive organic mixtures
gave an efficiency of 64%, surface tension negative systems an efficiency of 48%
(Ref. 5).) An indication of the HETP for laboratory column packings can be
obtained from the relevant manufacturers brochure. If it is proposed to use a
particular packing commercially, scale-up should be discussed with the
manufacturer.

An important practical feature of operating small diameter columns is to ensure
that heat losses are minimized. Columns must be adequately insulated and
preferably externally heated.

9

DESIGN USING ACTUAL PLANT DATA

There is quite often a need to uprate or debottleneck an existing plant column or
design a new plant for an existing product. Plant data may be available, but its
accuracy needs to be assessed and taken into account in any re-design.
Temperature and pressure indicators should be calibrated and if necessary flow
rates checked using a radio-isotope tracer method. Reported composition data
should be considered opposite the analytical and sampling procedures adopted.
If new plant data are required the A.I.Ch.E. Procedure for testing columns (Ref.
6) gives valuable guidance. In particular, every effort should be made to achieve
the measurement accuracy required to achieve the mass and energy balances
within the tolerances specified (± 3% for the mass balance, ± 5% for the
energy balance).
The normal procedure in simulation studies is to compare the plant data against
computed data obtained using a vapor-liquid equilibrium model. The basis of the
latter must be known. A good model will be based on experimental vapor liquid
equilibria data covering the relevant pressure conditions and important
composition regions. Alternatively it may be safe to assume ideal behavior, e.g.
for an homologous series of hydrocarbons. A poor, or approximate, model
could be considered to be one derived using estimation methods, e.g. UNIFAC,
especially for highly non-ideal systems.
If a good model is available then good agreement would be expected between
computed and plant data at the designated plant column plate efficiency or
packing HETP. Normally plate efficiencies will lie between 60% to 80%, may be
higher for a very close boiling mixture such as propylene-propane. Typically a
structured packing will contain about 2.5 theoretical plates per meter height, for
non-aqueous systems of relative volatility below 1.5.
If only an approximate model is available care must be taken in assigning a plate
or packing efficiency. Thus for processing reasons it may be required to replace
plates in a column with a structured packing. Reliance on an approximate model
to establish the number of theoretical trays in the existing column can lead to a
gross underestimation of the height of packing needed.


9.1

Uprating or Debottlenecking Exercises

Uprating or debottlenecking exercises can take several guises, ranging from the
need to compute tray hydraulics to considering major changes in still
configurations. If a good VLE model is available and gives good agreement with
plant data proposed changes can be confidently assessed using available
distillation and/or hydraulic programs. If only an approximate model is available a
common practice is to carry out plant trials, or maybe semi-technical
still simulations, to give confidence in the model and hence its reliability in
predicting the proposed changes.
Some aspects of uprating or debottlenecking exercises are presented in
Appendix C in the form of actual case histories.

10 REFERENCES

(1)

G E Keller, AIChE Annual meeting, Chicago, 1985.

(2)

GBHE-PEG-MAS-601- VLE Data : Selection & Use

(4)

W Featherstone, Department Paper R70/37, 'Azeotropic Systems: Quick
Method of Still Design' 4/3/70.
or
W Featherstone, Brit Chem Eng and Process Technology, 1121, Vol16,
No12, December (1971).

(5)

R K Badhwar, Department Paper 67/51, 'Distillation: Overall Efficiency of
Oldershaw Columns', 1/3/67.

(6)

Tray Distillation Columns: A Guide to Performance Evaluation A.I.Ch.E.
Equipment Testing Procedure, 1987.


APPENDIX A

WORKED EXAMPLE

Design a continuous distillation column to recover 99.9% wt/wt benzene from a
feed containing 40% wt/wt benzene, 40% wt/wt toluene, 20% wt/wt xylenes. The
feed rate is 20 te/h. The bottoms stream to contain < 0.1% wt/wt benzene (on a
toluene only basis).

A.1

ROUGH METHOD OF DESIGN

M.Wt
78.114
92.141
106.170

b.pt Feed
°C
80.1 Benzene
110.6 Toluene
139.1* Xylenes

% wt/wt

te/h

kmol/h

mole %

40.0
40.0
20.0
100.0

8.0
8.0
4.0
20.0

102.41
86.82
37.68
226.91

45.1
38.3
16.6
100.0

* m-xylene
The key separation is obviously (from the boiling points) that of benzene from
toluene.
Assuming ideality:

The α value is in reasonable agreement with that expected for components with a
boiling point difference of 30 °C.


Using the key component concept the feed contains 0.451 mole fraction of the
more volatile component (xF).
XD = mole fraction of more volatile component in distillate = 0.9991
XB = mole fraction of more volatile component in bottoms = 0.0008
i.e. 0.100 kmol benzene in bottoms or 0.1% wt/wt benzene in toluene.
Using Fenske:

Using Underwood, and as the distillate is of high purity:

For rough design purposes:
The number of theoretical plates = 2
×
The operating reflux ratio (R)
= 1.3 ×

Nmin
Rmin

=
=

28
1.8:1


Using the Fenske equation and substituting xF for xB:

Applying the Gilliland correlation the number of plates above the feed = 6.5 × 2 =
13.
Note:
The Gilliland correlation can be expressed numerically as follows:

The approximate reboil duty can be obtained by calculating (a) heat of
vaporization requirements, plus (b) sensible heat requirements.
(a)

Assuming a latent heat of 0.38 × 106 kJ/te, then

Heat of vaporization requirement = Distillate rate (8 te/h) × (R+1) × 0.38 × 106
kJ/te
= 8.5 x 106 kJ/h (ca 2.4 MW)


(b)

Assume specific heat capacity for organics
~0.5 cal/g °C (2.1 kJ/kg °C) then

Sensible heat requirement
= Bottoms rate × (Bottoms temperature - Feed temperature)
× 2.1 kJ/kg °C
= 12 te/h (133.6 - 100) °C × 2.1 kJ/kg °C × 103 kg/te
= 0.85 × 106 kJ/h (ca 0.24 MW)

Total heat requirement ~2.6 MW

A.2

DETAILED DESIGN OF COLUMN USING THE CHEMCAD
DISTILLATION PROGRAM

The benzene-toluene-xylenes system can be assumed ideal. The calculations
using CHEMCAD are therefore based on vapor pressure data and use the data
bank inherent in this distillation program.

A.2.1 Input Data
A tabular mass balance should be drawn up before using CHEMCAD. This gives
protection against unrealistic specification of product flows, particularly with more
complicated columns (multi-components, sidestreams, etc).
Mass Balance (kg/h)
Component
Feed
Benzene
Toluene
Xylenes

8000
8000
4000
20000

Distillate

Bottoms

7992
8
8000

8
7992
4000
12000


Approach:
(1)

Neglect xylenes in distillate.

(2)

Assuming bottoms toluene rate is 8000 kg/h, calculate benzene

(3)

Benzene in distillate by difference, i.e. 8000 - 8 = 7992

(4)

Total distillate from benzene specification:

i.e. toluene in distillate = 8 kg/h

(5)

Cross check the mass balance; all rows and columns should balance
correctly.

Convert this to the molar balance (kmol/h) and cross check.
Component
Benzene
Toluene
Xylenes

Molecular
Weight
78.114
92.141
106.170

Feed

Distillate

Bottoms

102.41
86.82
37.68
226.91

102.31
0.09
102.40

0.10
86.73
37.68
124.51

Thus the operation which it is required to simulate is outlined in Figure 2.


FIGURE 2

CHEMCAD SIMULATION

For a full explanation of the input data refer to the CHEMCAD user guide.


A.2.2 Output
Summary output:
Plate by plate outputs can be obtained if required.
A summary of the computed data is as follows.
Benzene Column
Summary of computed data
Distillate: 99.9% wt/wt benzene
Bottoms: 0.1% wt/wt benzene (on toluene only basis)
Computer
Output

Number of
Theoretical
Trays

Feed Tray
from top

Reflux
Ratio

Reboiler
Duty
MW

1
2
3
4
5

28
28
28
34
40

11
14 (mid-point)
17
17 (mid-point)
20 (mid-point)

2.3:1
1.9:1
2.2:1
1.6:1
1.5:1

3.1
2.8
3.0
2.5
2.4

The data show the importance of optimizing the feed tray position for a constant
number of trays. The effect of increasing the total number of trays opposite
decreased reflux or reboiler requirement is also exemplified The latter runs also
demonstrate that the reduction in R is decreasing as Rmin is approached. It
should be noted that a column designed too close to Rmin or Nmin will have little
scope to meet possible future requirements for purer product.
The computed data are in good agreement with those calculated using the rough
design procedure.


APPENDIX B

SLOPPY SEPARATIONS

B.1 EXAMPLE OF COLUMN MASS BALANCE FOR SLOPPY SEPARATION
Table 1 shows the component analyses of the feed, distillate and bottoms
streams of a naphtha prefractionator designed to remove components with less
than 6 carbon atoms as distillate. This is achieved as a sloppy separation in
which a range of the components in the feed appear in both distillate and
bottoms. The choice of which components are regarded as distributed is
not critical - it would be equally valid to regard 2 methyl pentane to benzene
(inclusive) as the distributed components. However, for analysis of plant data it is
preferable to choose as large a range as is reasonable, bearing in mind the
accuracy of analysis of low concentrations.
TABLE 1



TABLE 1



B.2

EXAMPLE OF CHEMCAD USE FOR A SLOPPY SEPARATION

A seven component feed is taken as an example of a sloppy separation. Table 2
shows the feed and partially completed mass balance.
The required separation is specified as:
(a)

overall recovery of components 4 & 5 to be 90%, and

(b)

total concentration of components 1, 2 & 3 to be 2 wt% max.

Both these specifications refer to the bottoms stream.

The bottoms rate, te/h was estimated as follows:
(1)

components 6 & 7 are unlikely to appear in the distillate and bottoms rate
= feed rate;

(2)

assume 90% recovery on each of components 4 & 5;

(3)

calculate sum of components 1, 2 & 3 to give 2 wt% in bottoms, i.e. 1.5
te/h.

To estimate the molar rate of components 1, 2 & 3 in the bottoms, an average
mol wt of 70 was assumed. By mass balance, the estimated distillate rate is 323
kmol/h - 325 was used in the CHEMCAD runs. An accurate estimate is not
necessary since the distillate rate is to be varied by CHEMCAD in meeting its
design specifications.

CHEMCAD input data and calculated liquid composition profiles for the total
reflux calculation. The number of stages used 25, was a guess. It can be seen
that the specification 2 wt% of components 1+2+3 is achieved between stages 5
& 6.
Stage Number
Wt% 1 + 2 + 3

:
:

5
3.9

6
1.5

Allowing 1 stage each for condenser and reboiler leaves 3-4 stages for the
column. ( input data and calculated product distributions to meet the column
specifications.) Heat duties and reflux ratio are also given. 8 stages were chosen
for this calculation, being approximately twice the minimum stages in the column
plus condenser and reboiler


APPENDIX C

SIMULATION USING PLANT DATA: CASE HISTORIES

C.1 UPRATING OR DEBOTTLENECKING EXERCISES: GOOD VLE MODEL
CASE A

BTX Still Train

There has been a requirement to uprate the BTX still train on A European
Aromatics Plant. The feed mixture containing benzene, toluene and C8 aromatics
can be assumed to behave ideally. The computed and plant data were in good
agreement (plate efficiencies of 65% to 75%).
The ability of the columns to operate at increased throughputs and/or higher
product specifications were readily checked using the appropriate tray hydraulic
and distillation programs.

C.2 UPRATING OR DEBOTTLENECKING EXERCISES: POOR VLE MODEL
CASE B

Combined Acetone Column

A fundamental change in the mode of operation on a European Phenol plant,
introduced in the early eighties. Conventional operation up to that time was to
recover crude acetone (containing water and cumene) from reactor product using
one of two stills designated to this duty. The crude acetone was further distilled to
produce refined acetone on a column situated on another plant.
The change involved combining the two existing stills on the plant in series to
recover refined acetone in one operation.
The separation basically involves the recovery of acetone from admixture with
phenol, cumene and water. The composition profile below the feed is such that
two liquid phases are present. The much used Wilson equation is not capable of
representing a system showing liquid-liquid immiscibility, but in practice this was
overcome by calculating Wilson E coefficients on the basis that an immiscible
system can be modeled by assuming it to be a very non-ideal miscible
system, i.e. an approximate VLE model.
A comparison of computed data using this model and actual plant data
nevertheless showed good agreement in terms of tops and bottoms compositions
and the column temperature profile.
The model was then used to predict the performance of one of the plant columns
operated at a higher reflux ratio such that a high purity acetone product could be

obtained. A plant trial, at reduced rates to allow operation at the high reflux ratio,
followed. The results were in agreement with predictions.
This gave the confidence required to use the model to predict the performance of
combining the two stills in series.
CASE C Ethylamines Distillation
A Methylamines plant was converted to produce higher amines, ethylamines
being the main products. At start-up production was limited to about 60%
flowsheet rate. This was a result of downcomer flooding and an unexpected
pinchpoint in the diethylamine-water system at 30 psig operating pressure versus
observations at 0 psig.
The VLE model of the highly non-ideal complex mixture of ethylamines-waterethanol was only approximate. However use of the model in a distillation program
indicated that a marked improvement would be expected if the sequence of
columns was changed. The conventional, existing still train removed
monoethylamine (MEA) overhead in one column, followed by diethylamine (DEA)
in the next (a direct sequence). The proposed still train was to recover MEA and
DEA overhead in one column followed by separation of these components on the
next column (an indirect sequence).
The indirect sequence was simulated on the semi-technical scale using
Oldershaw columns and results were in good agreement with predictions. Plant
trials followed prior to a full change to the new system. This allowed operation at
88% of flowsheet rate.

C.3

NEW PLANT FOR EXISTING PRODUCT

It would seem unlikely that there is not an adequate VLE model for an existing
still operation such that a new column could not be designed. This is not always
the case!
New Plant For Existing Product : Good VLE Model
There are no real problems if it is desired to build a new plant for an existing
product and a good VLE model is available. The predicted performance should
be in good agreement with existing plant data. The new plant design (maybe at
increased throughput and/or tighter product specification) can be based on model
predictions using an available distillation program and the subsequent use of
hydraulic programs.

However it is always worthwhile considering if the accepted still train
configuration is the most efficient.
CASE D :

p-Xylene 5 Sidestream Column

A European p-Xylene plant, as per previous units, continued the operation of two
columns to separate light and heavy-ends from isomerization product. Using a
model based on ideal behavior (the feed mixture contained benzene, toluene, C8
and C9 aromatics) it was shown that the operation could be carried out in one
still, the C8 aromatic product being recovered as a sidestream.
The combination of the two existing stills to operate as a sidestream column was
subsequently introduced onto the plant and gave considerable savings in energy.
New Plant For Existing Product : Poor VLE Model
CASE E :

New Diphenyl Oxide Still

There was a requirement to improve the quality of diphenyl oxide (DPO) product
with respect to dibenzfuran (DBF). The existing still contained Pall rings, the
proposal was to replace this with a column containing Kuhni Rombopak (a
structured packing).
The design of the new Diphenyl Oxide still was originally based on vapor
pressure data for DPO and DBF taken from a reliable source.
The data predicted sensible normal boiling points for the two components. On the
reasonable assumption that the system DPO-DBF behaves ideally, the vapor
pressure data gave a relative volatility (a) of 1.45 over the column operating
pressure range.
The distillation was simulated at Kuhni using Rombopak and the results showed
the vapor pressure data to be incorrect. The simulation results based on a = 1.45
indicated 6 theoretical plates per meter of packing - the expected performance
would be 2.5 to 3.5 theoretical plates per meter.
The vapor pressure data were therefore re-examined. The vapor pressure of
both components had been measured much earlier. These data predicted a
relative volatility of 2.0. Using this a value the data for Rombopak indicated 2.6
theoretical plates per meter – a more believable performance.


Comparing current plant performance against assumed values of relative
volatility confirmed that the best agreement was obtained with α = 2.
The design of the new still was therefore based on this value. The expected
performance was achieved when the new still was installed and operated.

C.4

POTENTIAL PITFALLS

Care must be taken in the extrapolation of VLE data.
(a)

Extrapolation beyond the composition range where VLE data were
measured. This is a trap quite easy to fall into, when you are designing for
a product with only 20 ppm impurity in it. If you have a non-ideal system
and the last data point was at 98% purity, much of your column will be
outside the range of the data and the K-value of the impurity
component could be wildly in error.

(b)

Extrapolation beyond the temperature range where the VLE data were
measured. This happened with an ideal system where the vapor pressure
data, from a reliable and renowned source, were extrapolated to low
pressure and predicted a relative volatility almost 30% in error - it is
described in Case E : New Diphenyl Oxide Still.

(c)

Extrapolation of plant data, especially where there may be some doubt
about the quality of the VLE model; thus the overall efficiency or HETP
fitted to the plant data may contain an element which reflects the errors in
the VLE model. Extrapolation to other conditions, especially involving
higher product purity or reduced reflux ratio, is likely to prove a failure.


DOCUMENTS REFERRED TO IN THIS PROCESS ENGINEERING GUIDE
This Process Engineering Guide makes reference to the following documents:

ENGINEERING GUIDES

GBHE-PEG-MAS-601

VLE Data : Selection and Use (referred to in
Clause 10, Ref. 2)

GBHE-PEG-MAS-602

Laboratory Distillation (referred to in Clause 8)

GBHE-PEG-MAS-603

Shortcut Methods of Distillation Design
(referred to in Clause 5)

GBHE-PEG-MAS-604

Distillation Sequences, Complex Columns and
Heat Integration (referred to in 7.2).


Design and Simulation of Continuous Distillation Columns

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Design and Simulation of Continuous Distillation Columns (20)

More from Gerard B. Hawkins (20)

Recently uploaded (20)

Design and Simulation of Continuous Distillation Columns