Professional Documents
Culture Documents
a,
Pnganron Press Ud.
COLUMNS
BODO LINNHOFF*
Department
of Chemical
Engineering,
University
of Manchester
Institute
and
Department
ROBIN SMITH
of Chemical Engineering, Bradford University, Bradford, West Yorkshire,
(Received
26 November
England
1982)
Abstract-There
are many well-known schemes for better energy efficiency in distillation. Examples are thermal
coupling, multiple effect, heat pumping, etc. Usually these schemes are discussed for individual columns in
isolation, independently from the overall process they are a part of.
This paper puts the design of individual distillation columns into context with the heat integration for the overall
process. An insight is discussed which leads to interesting conclusions. Firstly it is shown that good integration
between distillation and the overall process can result in columns operating al effectively zero utility cost. Secondly it
is shown that some of the traditionally well-known schemes referred to above may in certain circumstances turn
out counter-productive.
They can prevent good integration and therefore eliminate the change of obtaining zero
utility costs.
Generally, the paper defines good integration as a column not crossing the heat recovery pinch of the process and
either the reboiler or the condenser beine inteerated with the process. If these criteria can be met, energy costs for
distillation can effectively be zero.
INIRODUCTION
should be addressed.
1175
1176
B. LINNHOFF et al.
Feed1
25oc
I II
1.
FEEDTO
10.0
25
190
WthSlpy
Data
H
u'
145
146
190
0
9'::
10.0
2.
RZYCLE
3.3
-20
200
3.
FUJI3
2.4
480
160
3.5
250
173
2.52
140
80
8.0
140
35
GAS
FmFUP?ACE
4.
REXXCR
EXIT
SIRFAM
5.
EETIJ_lS?mSTR~
FRZMCUEMX
6.
FmmTD
DISTILIATI~
Tram
EnthalW
T
140
110
35
7.axJ_wN1
RFBXLER
1.5
6.
C0IJJY.M
1
_FZ3SER
1.6
alm4w 2
REBDILXR
3.0
10. CorJ.l?m2
CCWDF?J!TER
3.1
9.
Data
8.0
2.0
0
100
liq
-20
100
VaP
-20
vap
140
li
lis
140
vap
65
vap
65
liq
Heat integration
1177
RECOVEXY
PINCH-A
Rl3vIEw
The following review of heat exchanger network principles contains only as much detail as is required in the
present context. The principles are discussed in detail in
other references&101.
The pinch
Any chemical process can be considered as a series of
streams requiring heating and cooling. This heating and
cooling is achieved by a combination of process interchangers (matches between process streams) and heaters
1178
B.
LINNH~FF
et al.
SOURCE
Fig.
3.
process below the pinch, this heat would then add to the
external cooling load. In other words, any heat transferred across the pinch must lead to an equal increase in the
consumption
of both hot and cold utilities.
By equivalent
arguments
any heat rejected to cold
utility at temperatures
above the pinch and any heating
with hot utility below the pinch roust increase utility
requirements [9].
Utility pinches
The heat flow cascade also allows the optimum levels
of alternative utilities to be determined.
In Fig. 3 all hot
utility is supplied at a temperature
higher than the
highest temperature
within the process. Often, however,
not all the heat is required at such a high level and it is
more economical to make use of more moderate temperature heat sources, such as low pressure steam instead of high pressure steam. Heat can be introduced at
lower levels above the pinch as long as the heat flows
remain positive.
If we maximise
the use of more
moderate levels we may reduce heat flows to zero as in
Fig. 4 and so introduce further pinches. We call these
pinches utility pinches to distinguish
them from the
original process pinch.
Below the pinch the situation is analogous. Often, we
can use reject heat to, far example, raise steam rather
than cooling water and if the use of the higher level sink
is maximised we may create a utility pinch.
TEMP.
SOURCE
PINCH
PINCH
Qs
QCOll
Fig. 4. Multiple
utilities.
Q2
Unpinched
processes
A somewhat special process cascade is shown in Fig.
5. Here no heat at all is rejected to cold utility and we
use utility simply to maintain the enthalpy balance of the
process. The only zero heat flow is that out of the lowest
temperature
interval. This zero heat flow does not indicate a pinch situation but simply the fact that cold
utility is not required. The process does not have a pinch.
We refer to problems such as this as threshold problems
and they are discussed in more detail by Linnhoff and
Hindmarsh[9]. The name stems from the characteristic
LOWER
HEAT
UTlLlTY
Fig.
5. Threshold
Q3
problems.
1179
Ohmin+
0r.b
o,-a.,
4cv.b
Heat
Fig. 7. Dislillation across the pinch.
4-l
01
lkz..i4..dl
0-b
-v*-a,.aFeed
Heat
in
B. LINNHOFF
et al.
1180
Mukiple
pinches
Unless multiple utility levels are used there is usually
only one pinch point in the heat flow cascade, i.e. one
point with exactly zero heat flow. However, in practice it
is possible for there to be several points where the heat
flow is too small to allow integration of distillation
columns because of the heat load limit discussed above.
These points effectively act as pinches as far as integration of distillation columns is concerned.
If a
column lies between two such near-pinches
then both
condenser and reboiler must be integrated.
SINK
Fig.
10. Heat
1Qhmin- Qcond
+ Qcmin
QrDb
Heat
integration
of distillation
AWAY
Pressure changes
The pressure chosen to operate a distillation column
influences many important design parameters, e.g. relative volatility, vapour density, shell thickness,
etc.
However its most important influence in the present
context is in determining the condenser and reboiler temperatures, and hence the levels of heating and cooling
required. These temperatures are crucial as they determine the position of the column relative to the pinch. If
they span the pinch, one option is to change the column
pressure. We can in principle either increase or decrease
the pressure, thus changing the columns position relative
to the pinch. However in practice there are limits.
Raising the pressure. Here we aim to integrate the
column condenser by lifting it above the pinch. The
separation will generally become more difficult (the relative volatility decreases) requiring either more plates or a
larger reflux ratio. However, the latent heat of vaporisation decreases,
compensating
to some extent for the
PINCH
01
PINCb
Utility
Rrfrigcratlon
Fig.
12. Integration
of column
1181
with refrigerated
condenser.
B. LINNI~~FF et al.
1182
*_QQd
I
Feed
Qhmin
+ IOrrb,
-Ocona,)
PI
Qr.bz
pt
I.
0
pf
Or.bz<Q,-=&rb
PINCH
QWb2
or2
O,-Qr*a
eandl
Side - stream
Rcctliicr
1 ohmin +
PINCH
Q+
1183
oh,,, + (Oh.-
0,
t
horizontal thermosyphon
and vertical thermosyphon.
The first three employ boiling on the outside of tubes.
Mixing occurs as the liquid boils such that all of the heat
(including any for preheating the subcooled liquid) must
be supplied above the dew point of the vapour leaving
the reboiler. If a vertical thermosyphon is used, however,
the liquid boils in the tubes and a significant temperature
gradient can occur along the tubes. This can in principle
be matched with another fluid flowing counter-currently
such that the temperature driving force keeps benefitting.
Usually, however, this would only be justified if heat was
transferred
at low temperature differences
and if the
boiling range was large. It is unlikely that this would ever
be attractive except in low temperature processes. In
general studies, we recommend to assume that all heat
must be supplied above the dew point temperature of the
vapour.
Consider now condensers. The same basic arguments
apply. When a multicomponent vapour condenses it does
so over a range of temperatures. Whether or not we can
exploit this depends on our choice of heat exchanger. If,
for example, we choose in-tube vertical condensation
then we can exploit the condensation curve when heating
a process fluid. If, however, we choose horizontal shellside condensation
then all heat must be removed at a
temperature lower than the bubble point of the liquid.
Again, practical applications
where the condensation
range is exploited for belter driving forces will be few
and far between outside the low temperature
field.
Throughout the present paper it is assumed that heat is
removed in condensers at a constant temperature lower
than the bubble point of the overhead liquid.
Does the pinch change?
So far in our analysis of where to place distillation
columns in relation to the process we have assumed that,
once the column has been integrated into the cascade,
the position of the pinch will not change. In principle, the
1184
B.
L~NNHOPF
et al.
It is now appropriate
to return to the flowsheet
presented in Figs. 1 and 2. As stated earlier, the process
in Fig. 2 is optimal as far as heat recovery is concerned.
Conventional heat recovery procedures confirm that the
energy target has been achieved. However, the procedures also identify a pinch temperature of 135C. This
pinpoints that Column 2 is operating across the pinch.
After this realisation, we can improve our energy
performance further provided we shift Column 2 away
from the pinch. First, we consider changes to the operating
pressure. A very large increase in pressure would be
necessary to place the column wholly above the pinch (i.e.
condenser temperature > 135C). Lowering the pressure,
however, soon achieves appropriate placement of the
reboiler. The cascade identifies the pressure change
necessary so that we can accommodate the reboiler load of
3.0 MW. A reboiler temperature
lower than 110C is
required. Assuming a similar shift in condenser temperature, 65-3X, we can still use cooling water at the top
of the column. An alternative would be a smaller reduction
in pressure along with splitting the reboiler load via an
intermediate reboiler. This would have been worth considering if the pressure change had resulted in a requirement for refrigeration.
This analysis is independent of actual network design.
It provides an indication, prior to design, of what can be
achieved. At a second stage, actual network design studies (using the Pinch Design Method[9])
with the
appropriate pressure adjustment in Column 2, give the
design in Fig. 16. The steam usage for the total flowsheet
is now such that the front-end
plant plus the distillation operates with no more steam than the frontend alone, contradicting the belief that distillation is an
expensive operation in energy terms! By appropriate
integration distillation can require zero utility.
The design in Fig. 16 shows one other change from the
original flowsheet in Fig. 1. In Pig. 1, the condenser on
Column 1 is refrigerated and the column feed is watercooled with the cooling effect being limited by the cooling water temperature. However, vapout leaves the top
of the column at -20C. No use is made of this vapour in
Fig. 1, but in Fig. 16 it is used to cool the feed beyond
CASE STUDY
1185
BOC:+
Feed
UKlK
ohmi,
A.
iw +
OC.d - Orrb)
w + (Qcond-Qd
Od3
Cd"
Feed
H.R
W
PIHCH
QC0.d
1186
B.
LMNHOFF
et al.
within the two columns. We therefore lose the flexibility of choosing different pressures within different
columns to suit integration. Along with this it should be
noted that the heating and cooling duties that remain are
at the most extreme temperature levels. Thus it is possible to envisage the situation shown in Fig. 19 where
thermal coupling actually prohibits good integration. The
conventional arrangement of columns has a larger total
load but the pressures can be chosen independently such
that neither column operates across the pinch. The
thermally coupled arrangement,
on the other hand,
with its wider temperature range and uniform pressure in
both columns, may have to span the pinch. Clearly in
these circumstances thermal coupling is counter-productive resulting in a higher overall utility usage.
We conclude that thermally coupled systems, like heat
pumps and multiple effect schemes, should not be considered in isolation but only after the overall process has
been analysed in terms of the heat flow cascade.
Intermediate
Thermally
coupled columns
Returning to Fig. 14, the use of
advocated as a means of reducing
saving energy. However another
coupled arrangements is the need
- Fe1
PREWOVS
WORK
11
P
Cd
fi
Ohmin
*____I
PINCH
PINCt
-
t aem,*+
Qcona,
Fig.
18. Multiple
effect
distillation: dont
tegration studies.
use it prior
to in-
+ Qcmda
Heat integration
of distillation
1187
Individual columns
The present
paper discusses the design of individual
distillation columns in the context of an overall process.
This is important as quite clearly distillation columns do
not usually operate in isolation. Any strict division between distillation columns and the rest of the process is
not one that practising engineers will agree with.
The paper shows that, while it is important for the
engineer to be aware of techniques such as double-effect
distillation, heat pumps, etc. it is equally important that
he appreciates the overall process. Optimising the design
of individual columns in isolation from the rest of the
process
may sometimes
have a counter-productive
effect, spoiling opportunities
later for good integration
and therefore adversely affecting the performance of the
overall process.
On first sight, there seems to be some similarity between the present paper and the work of Umeda et al. [5].
Like the present paper, Umeda et a?. discuss the design
of distillation columns against the background of a wider
heat integration task. (Their background process simply happens to be a distillation train only.) Further,
Umeda et al. quote the pinch concept. However, they do
not make any use of the pinch concept in design. Rather,
they offer a number of heuristic rules for the evolution of
a design. These rules are to be applied on a trial-anderror basis and might or might not improve overall
performance.
The importance of the correct placement
of a column relative to the pinch remains unrecognized
by Umeda et al.
Separation trains
Acknowledgements-Most
work described in this paper was carried out while all three authors were in the employment of ICI.
The authors are grateful to ICI for permission to publish this
paper. Also, they much appreciate the helpful comments from
the referees.
the previous
NOMENCLATURE
H enthalpy, kW
pressure, bars
5 heat flow, kW
Qh hot utility supplied, kW
QC cold utility required, kW
Qhmin, Qcmin
kW
REFERENCES
Proc. 197554
1188
B. LINNHOFFef al.