H E AT T R A N S F E R

Dont Let the

Pinch Pinch You
Blind obedience to
pinch technology
rules and
procedures can lead
to unnecessarily
complex and
expensive, and
potentially
hazardous, plants.
Here we show how
plant structures
that are both simple
and energy efficient
can be generated.

G. T. Polley and P. J. Heggs,

Univ. of Manchester Institute
of Science and Technology

t has now been more than 20 years

since the pioneering work of Umeda
and his coworkers at Chiyoda laid the
foundations of modern process integration technology (1). Over the fouryear period of 19781982, this team introduced the concept of overall process
system analysis, reintroduced the concept
of composite curves and showed how
these could be used to determine the utility needs of a process, developed a synthesis strategy based upon overall network costs, identified the heat recovery
pinch for the first time, demonstrated
how design of the core process and the
heat-recovery system could be linked
through manipulation of streams around
the pinch point, and introduced the heat
demand and supply (HDS) diagram (now
more commonly referred to as the grand
composite curve) and the raffinate diagram (now used in overall site analysis)
(14). An overview of the technology is
now available on the Internet (at
www.Pinchtechnology.com).
The methodology was subsequently
extended and popularized, most notably
through the development of the pinch
design method (which moved network
synthesis away from the computer and
into the hands of practical engineers) and
through the introduction of the powerful
concept of target before design. With
these later developments, the process
pinch became the pivotal point in heatrecovery network design, and the pinch
has now become the focus of the methodology. (The method has even become
known as Pinch Technology.)

The Chiyoda team did not place as

much emphasis on the pinch as is done
today. To them, the pinch was important
in directing process modification rather
than controlling design. They had valid
reasons for this philosophy. They observed that, although it is possible to construct composite lines for every heat sink
or source stream involved in the processing system, the results of heat integration
give complex integrated systems, which
may cause difficulty in operation and that
final heat integration between reaction
and separation systems is carried out by
engineering judgment of appropriate solutions through practical considerations.
Their message could be interpreted as:
Keep it simple. They achieved this by
adding an extra element to their synthesis
strategy the systematic addition and
subtraction of streams to and from the
analysis.
Failure to observe this important message may have cost dearly. In 1990, an
engineer from a major oil company complained to one of us that integration was
going too far. Flowsheets were becoming
too complicated. Later, he commented
that integration had reached such a level
that some plant operators had difficulty
understanding where individual streams
were going.
This article demonstrates a strategy
based on the keep it simple message
that results in simpler (and cheaper) integration schemes (5). With this strategy,
the overall process pinch becomes less
important different regions of the process have their own pinch points that can

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999 Copyright 1999 American Institute of Chemical Engineers. All rights reserved. Copying and downloading permitted with restrictions.

H E AT T R A N S F E R

Figure 1.
Initial network design
for an aromatics plant.

40
160

83.4
C

119

220

327

10

4.34

220

60

79

C
1.14
60
103.5
C

149.9
2

215.8

189

220

159.5

160

17.39

100

201
85.9

35
3.56

85

141

10.1

3.86

1.61

141

6.6

Source: Redrawn from (6).

(1)

The installed cost is about 3.5

times higher than the capital cost.
This suggests a significant economy of scale. For instance, the fixed
cost element itself would provide for

0.2

190.3 243.8

189
9.6

an additional 40 m2 of surface in a
unit of 200 m2 nominal size. Since in
network optimization it is often possible to compensate for the removal
of one unit by adding additional surface elsewhere, the designer should
seek to eliminate small exchangers
from the network in favor of large
ones.
Figure 2 shows how much additional surface can be purchased in another single exchanger with the

Area Added to Existing Exchanger, m2

C = 4,600 +

920A0.7

138
170

140

Problems with the

pinch design method
The pinch design method has a
number of weaknesses. These can be
demonstrated using an example from
the literature. Figure 1 is the initial
network derived by Ahmad and
Linnhoff for a simplified aromatics
plant (6).
A recent study (7) indicates that
the capital cost (C) of a typical single
heat exchanger (TEMA Type AEM,
5-in. tube, 10-bar design pressure,
etc.) is related to the total heat-transfer surface area (A) by the following
equation:

9.9 164

18.55

60

differ from that for the overall process. The pinch point resumes its role
as a guide to process changes rather
than acting as the point underpinning
the design.

300

300

H
0.25

10.7 11.25

money saved by removing an exchanger. The unit to be removed is

represented by the four lines labeled
A = 0 through A = 100 m2; the x-axis
refers to the size of the existing unit
to which extra area is to be added,
and the y-axis refers to the amount of
extra area that can be purchased.
So, if a proposed 50-m2 exchanger is
removed from the design, approximately 140 m2 could be added to another existing exchanger currently

300

A = 100 m2

A = Area of
Exchanger
Removed

250
200

A = 50 m2

150

A = 25 m2

100

A=0

50
0
25

50

100

150

200

250

300

350

400

450

Area of Existing Exchanger to Which More Surface is Added, m2

Figure 2. Area purchased through removal of exchangers of various sizes.

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

500

Reactor
1

Gas
Separator
1

Distillation
Column 1

5
4
Distillation
Column 2

Gas
Separator
2

Reactor
2

7
Unit E
9
Source: Redrawn from (6).

Figure 3. Process flow diagram for aromatics plant.

sized at 100 m2 (making the new unit
240 m2), or about 180 m2 could be
added to an exchanger currently sized
at 300 m2 (making it 480 m2), before
the purchase cost of the exchangers
for the network is increased. This
analysis considers only the exchanger
purchase cost; no credit has been assumed for cost savings associated
with piping and other auxiliary equipment. Even so, a strong capital cost
trade-off is indicated.
The pinch design method divides a
design optimization problem into two
parts at the pinch and produces designs for each subnetwork before
merging the results to provide an initial structure for subsequent optimization. By using the check-off heuristic,
each subnetwork provides a design
that uses the minimum number of
units. However, when these designs
are merged, the result is an initial design that uses more than the minimum number of units for the overall
problem.
For the aromatics plant discussed
in Ref. 6, the minimum number of
units required for the network is 10.
The number in the initial design is

15. This suggests that, if installation

costs are high, this initial design is a
long way from the optimum configuration. So, lets look at how some of
the exchangers in this design can be
removed.
Consider Exchanger 8. This is one
of the smaller exchangers in the network and could be a candidate for removal. However, it cannot be removed. Stream 3 must be taken to a
temperature of 164C. The only other
recovery unit on this stream is Exchanger 3. The highest hot-stream
inlet temperature possible for this
unit is 160C.
Now consider Cooler 2. This, too,
is one of the smaller units. This unit
cannot be removed because the
neighboring heat-recovery exchanger
links with a cold stream having a supply temperature that is identical with
the required outlet temperature from
Cooler 2.
This raises the question: Can any
of the coolers be removed?
The required outlet from Cooler 3
is below that of the cold stream entering Exchanger 2. So, this cooler cannot be removed.

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

The required outlet from Cooler 1

is just 5 above that of the cold
stream entering Exchanger 1. Not
only is this close to the practical limit
for a shell-and-tube exchanger, but it
would probably result in quite a high
capital cost penalty, making the removal of Cooler 1 uneconomic.
This examination indicates that the
removal of small uneconomic exchangers can be difficult to achieve
and that the initial design derived
using the pinch design method can result in nonoptimal designs.
Another weakness of the methodology is its failure to account for the
nature of the streams being processed
for example, two-phase streams.
Not only is the transport of twophase streams problematic and expensive (requiring large-diameter
piping), but, given the difficulty of
maintaining good mixing of the
phases, the number of heat exchangers in which such a stream is used
should be minimized.
Figure 3 is the process flow diagram (PFD) for the aromatics plant.
This diagram suggests that Streams 2
and 7 are two-phase streams. In the
initial design presented in Ref. 6,
Stream 2 goes through Exchangers
10, 5, and 1 and through Cooler 2,
and Stream 7 passes through Exchangers 6, 3, and 2 and through
Cooler 3.
Failure to relate network structure
to the PFD has many implications.
This becomes apparent when one
compares the proposed initial structure and the PFD.
Consider Stream 4. This stream
comes from the base of first distillation column and forms the feed to the
second reactor. With this network design, Stream 4 is first used to extract
heat from the feed to the second gas
separator. It is then matched with the
product from the second reactor before being used to recover heat from
the stream leaving the base of the
second distillation column. Finally,
before at last being raised to the necessary temperature and being fed to
the second reactor, it is matched with

H E AT T R A N S F E R

Figure 4.
Three simple
subnetworks.
C

1
a.

b.

2
1

c.

the product from the first reactor.

Now consider Stream 3, which issues from the first gas separator. It is
first matched with the product from
the first reactor before being transferred to the area of the second gas
separator, where it is matched with
separator feed. Finally, it is matched
with the bottoms product of the sec-

ond distillation column before being

fed to the first distillation column.
Failure to recognize the impacts
of network design on plant piping
and process flows is serious. It not
only significantly increases piping
cost penalties, but it also has safety
implications.
Finally, consider the software
Figure 5.

First Division
Group 1
Stream 1
(TAC)
Second Division

Streams 2

Group 1
Streams 1
2
(TAC)
Third Division
Group 1
Stream 1
2
(TAC)

Streams 3

Streams 4

Group 2
5
6
(TAC)

5
6
(TAC)

5
6
(TAC)

And so on...

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

Sequential
decomposition
of a process
integration
analysis.

needs associated with the existing

technology. Two types of programs
are needed a targeting program
and a network optimizer; in todays
computer-orientated environment, it
could be argued that a network design
program is needed as well. This is a
lot of software, some of which is sophisticated. Yet, as already pointed
out, the end result (despite the comfort provided by sophisticated optimization) can be a nonoptimal process design.

Observations regarding
network design
Networks that contain the minimum number of units generally have
a quite simple structure that can be
broken down into a number of even
simpler self-contained subnetworks.
Typical examples of such structures
are shown in Figure 4.
The structure in Figure 4a consists
of just one heat-recovery exchanger
and a heater. Full use is made of
available process heat. There is no
scope for capital cost saving, because
in terms of duty, the structure is already optimal.
The structure in Figure 4b consists
of a single heat-recovery unit, a
heater, and a cooler. Varying the size
of the heat-recovery unit affects both
the capital and energy costs of the
system, which poses an optimization
problem. However, it is a straightforward problem that can be solved by
applying integration range targeting
to the two-stream problem. A sophisticated optimization program is not
required.
The structure in Figure 4c has
three process streams and two heatrecovery units plus one heater and
one cooler. Here, too, an optimization
problem exists. However, the structure can be decomposed into two
parts: Exchanger A, which as a standalone unit does not need to be optimized, and a structure similar to that
already considered (in Figure 4b),
which can be optimized using a targeting program.
Amidpour and Polley (5) intro-

 Figure 6.
Process flow diagram
for the aromatics
plant in the example.

1
2

R1

T1

T2

T3

T6

13

T4

10
12

11
14

6
16

T5

17

R2

20

22

T10

T7

18

15

duced problem decomposition into

process integration analysis. In this
procedure, an overall integration
problem is decomposed into a number of self-contained zones in accordance with the PFD. Integration is
then restricted to streams within
zones unless integration between
zones was clearly economic. In another paper (8), they showed how
piping constraints could be dealt with
by ensuring that the more-difficult
streams (such as two-phase streams,
condensing vapors, evaporating liquids, and low-pressure gases) are not
transferred out of their zones.
By decomposing the overall problem in such a manner, the engineer
ends up designing networks that are
local in terms of operation, which
therefore can easily be understood by
operators. They are also associated
with local unit operations and thus
are unlikely to result in unnecessarily
expensive pipe runs. Finally, rather
than dealing with a single large net-

21

19

work-design problem, the engineer is

handling a small number of subproblems dealing with fewer streams. The
design of these smaller individual
networks is likely to be a simpler task
than developing a network for the
whole process. The designer will
often be able to solve the series of
subproblems faster, and to better effect, than the overall problem.

Problem
decomposition procedure
The following procedure for problem decomposition (Figure 5) can be
applied using simple process-integration targeting software:
1. Number the streams sequentially.
2. Apply range targeting to the full
stream set and establish a reference
total annual cost (TAC).
3. Split the stream set into two
parts, the first set consisting of
Stream 1 and the second set consisting of the remaining streams (here,
Streams 2 to 9).

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

4. Apply range targeting to both

sets and sum the results.
5. Derive new sets by changing
the position of the boundary, so that
the new first set contains Streams 1
and 2 and the new second set contains the remaining streams (Streams
3 to 9).
6. Repeat Steps 4 and 5 until all of
the boundary positions have been
examined.

Stagewise application
of problem decomposition
The order in which the streams are
listed affects the results of the decomposition analysis. By arranging them
in the order in which they appear on
the PFD, the designer can identify
cost-effective local integration (as opposed to cross-plant integration). This
has benefits in terms of piping cost,
plant operability, and plant simplicity.
But, once these opportunities are exhausted, this ordering has no benefit.
For general thermal integration,

H E AT T R A N S F E R

the streams should be ordered in line

with the Ponton-Donaldson Heuristic
of: Match the hot stream having
the highest supply temperature with
the cold stream having the highest
target temperature. (9)
In applying this approach, there
will be occasions where the residual of the hot stream (i.e., the hot
stream condition after a match has
been made) may be important. So, the
analysis is conducted in the following
stages:
1. Order the streams in line with
the PFD and identify the local integration opportunities.
2. Having exhausted the opportunities for cost-effective local integration, reorder the remaining streams in
line with the Ponton-Donaldson
Heuristic and repeat the analysis.
3. If Step 2 results in significant
potential cost penalties, identify the
hot stream residuals that influence the
heat recovery and subdivide these hot
streams (still ordering in line with the
Ponton-Donaldson Heuristic).

Overall ordering of a
process integration study
A process integration analysis
using this approach involves the following five stages.
1. Problem simplification removal of streams from problem.
2. Identification of process
changes improved overall efficiency through energy and utility cost
savings, and capital cost reduction.
3. Setting of final problem selection of utility levels, specification
of utility costs, selection of heattransfer parameters, and adjustment
of stream ordering, if desirable.
4. Decomposition analysis on the
basis of the layout identification of
cost-effective local integration.
5. Decomposition analysis on a
thermal basis development of the
remainder of the network.
Example:
aromatics solvent plant
Figure 6 is the PFD for an aromatics solvent plant. We will use it to

Table 1. Stream data for the example.

Stream
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

Name

CP, kW/C

Make-up
T1 Tops
T1 Bottoms
T2 Reboiler
T2 Condenser
T2 Bottoms
T3 Feed
T6 Reboiler
T6 Condenser
Raffinate
T4 Preheat
T4 Reboiler
T4 Condenser
T4 Bottoms
T5 Reboiler
T5 Condenser
R2 Feed
R2 Effluent
T10 Reboiler
T10 Condenser
T7 Reboiler
T7 Condenser

8
72
30
12,000
12,400
14
46
7,600
5,900
46
16
4,100
3,500
18
8,000
9,500
75.2
52.8
2,200
2,000
3,800
3,460

demonstrate the basic analysis technique. Consequently, we will not

cover some of the more complex issues such as pressure drop specification, consideration of exchanger technology, and so on. The problem is
therefore defined in terms of stream
heat capacity flow rates, and physical
property information is not required.
The streams have been listed in the
order in which they appear on the
PFD. The stream data (heat capacity
flow rate (CP), supply temperature
(Ts), target temperature (Tt), and load)
are summarized in Table 1.

Stage 1
Simplify the problem
The stream data are entered into a
process-integration targeting program. (We have used INTEGRITY,
developed by ESDU, which already
incorporates the problem decomposition procedure.)

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

Ts, C

Tt, C

Load, kW

20
38
38
140
66
140
65
79
56
80
38
120
59
121
110
59
88
188
66
48
114
50

80
80
49
141
65
30
38
80
55
30
80
121
58
38
111
58
204
8
67
47
115
49

480
3,024
330
12,000
+12,400
+1,540
+1,242
7,600
+5,900
+2,300
672
4,100
+3,500
+1,494
8,000
+9,500
7,564
+5,228
2,200
+2,000
3,800
+3,46

The hot utility temperature is set at

500C, and the cold utility temperature is set at 20C. A point target, at a
minimum temperature approach of
10C, is generated. Examination of
the HDS diagram (Figure 7) shows
the following:
one overhead condenser is below
but close to the pinch (examining the
temperatures and loads in Table 1,
this is identified as overheads on T2);
two reboilers are above but quite
close to the pinch (again using the
data in Table 1, these are identified as
the reboilers on T6 and T10);
the remaining reboilers are well
above the pinch and, given the shape
of the HDS diagram, which shows no
high-temperature heat surplus, these
cannot be driven through heat recovery; and
the remaining condensers are
well below the pinch and, given that
the HDS diagram indicates no low-

350

300

Internal Temperature , C

250

200

T2

150
T5
100

T7

T4

T6
T10
T2

50

T5

T4

T7

T10

0
0

5,000

10,000

15,000

20,000

25,000

30,000

35,000

Enthalpy, kW
Minimum temperature difference = 10C

Figure 7. Heat demand and supply diagram for the aromatics plant in the example.
temperature deficits, heat recovery
from these units is not justified.
Therefore, assume that the reboilers on Columns T2, T4, T5, and T7
are driven by hot utility, and delete
these streams from the problem. Also
assume that the condensers on
Columns T4, T5, T6, T7, and T10 are
cooled using cold utility, and delete
these streams from the problem.

Stage 2
Identify process changes
The condenser on Column T2 is
below and close to the pinch while
the reboilers on Columns T6 and T10
are above and close to the pinch. In
addition, the load on the condenser is
close to the combined reboiler loads.
The pressure at which Column T2
operates can be increased. This would
result in an increase in the temperature at which the overheads condense.
It may be possible to use this vapor to
drive the two reboilers. If the temperature of the overheads could be raised
high enough, very-low-pressure

steam could be generated and this

used to drive the reboilers. An alternative approach would be to reduce
the pressures on Columns T6 and
T10. Evaluation of these alternatives
requires the use of a process simulator and is outside the scope of this
article.
If process modification is not viable, the HDS diagram indicates that
the condenser should be cooled using
cold utility and the reboilers driven
using hot utility.
The following actions are therefore recommended:
1. Examine possible process modifications. (Since the alternative is to
run both units using utilities, this can
be done after the rest of the heat-recovery system has been designed.)
2. Delete all of the condensers and
reboilers from the problem.

Stage 3
Set the final problem
Utility levels and costs. Examination of the HDS diagram suggests

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

that the best form of hot utility is

steam. Heat must be provided up to a
temperature of 204C. This suggests
a steam temperature of around
220C.
The hot utility cost is then calculated as follows. Cost of fuel oil =
$0.16/L; potential heat extraction
from fuel = 40.8 MJ/L; base cost (including 10% to account for capital
costs and assuming that the fuel is
used at 85% efficiency) =
1.294(0.16)/40.8 = $0.0051/MJ. Temperature factor (which accounts for
the variation in the heat of vaporization of steam with temperature) =
1.043 + (2.407 10-4)(220) (5.26
10-6)(220)2 = 0.844. Steam cost =
0.0051/0.844 = $0.00604/MJ =
$0.0218/kWh.
Examination of the HDS diagram
indicates that cooling water (at 20C)
is a suitable cold utility. The cooling
water cost is assumed to be
$0.0016/kWh.
Capital costs assumptions. The
capital cost of an exchanger is given
by:
C = a + bAc

(2)

The values of the constants a, b,

and c for a typical carbon steel exchanger are given in (7). The values
used here are a = 16,000, b = 3,200,
and c = 0.7.
Annualization factors. The following annualization factors are assumed: 8,000 h/yr operation, 10-yr
plant life, and 10% interest rate.
Heat exchanger parameters.
Multipass exchangers are assumed,
and the X factor (that is, the limit
placed on the approach to maximum
thermal effectiveness (10)) is set at
0.9. The maximum exchanger size is
set at 500 m2/shell.
Typical stream heat-transfer coefficients and fouling resistances are
determined (for example, using Ref.
11). For liquid organic (light)
streams, Ref. 7 suggests approximate
film heat-transfer coefficients of
1,500 W/m2K and a fouling resistance of 0.0004 m2K /W. Combining

H E AT T R A N S F E R

the two resistances and then inverting

yields a duty coefficient of 937.5
W/ m2K.
For steam, a coefficient of 7,000
W/ m2K is recommended with a
fouling resistance of 0.0001 m2K /W.
Combining gives a duty coefficient of
4,120 W/ m2K, so a value of 4,000
W m2K will be used. For cooling
water, a coefficient of 4,000 W/ m2K
is suggested with a fouling resistance
of 0.0002 m2K /W. Combining gives
a duty coefficient of 2,220 W/m2K
m2K, so a value of 2,000 W/ m2K
will be used.

Stage 4 Decomposition
based on layout
The streams are ordered in accordance with their appearance on the
PFD. A range target for the full problem is then determined. The minimum TAC is found to be
$581,000/yr. This is a theoretical
minimum cost. The TAC of the final
design can be expected to be 1015%
higher than this.
Range targeting can now be applied systematically to different data
sets to identify how the problem can
be cost-effectively decomposed into
local integration problems.
Decomposition into two parts, one
group containing Streams 1 to 8 and
one group containing Streams 9 and
10, yields a system with a theoretical
minimum TAC of $611,000/yr. This
is made up of a theoretical cost of
$194,000/yr (at an optimum approach
temperature difference of 15C) for
the first group and an actual optimized cost of $417,000/yr (at an optimum temperature difference of 10C)
for the heat-recovery system involving the reactor streams (Streams 9
and 10).
Comparing this decomposed cost
of $611,000/yr with the theoretical
minimum of $581,000/yr indicates a
potential cost penalty associated with
this real match of $30,000/yr. This is
5.2 % of the theoretical minimum
cost.
Separate the reactor feed stream
(Stream 9) and reactor effluent stream

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

Table 2. Remaining five streams to undergo

decomposition on a thermal basis.
Stream

Ts, C

Tt, C

38
38
140
65
80

80
49
30
38
30

T1 Tops
T1 Bottoms
T2 Bottoms
T3 Feed
Raffinate

Load, kW
3,024
330
1,540
1,242
2,300

Table 3. Summary of final design for the example.

Local integration around reactor
TAC = $147,000/yr
Potential Penalty = $3,000/yr
Heat-recovery match (reactor feed against reactor effluent) requires 714 m2
Heater on reactor feed area = 114 m2
Cooler on reactor effluent area = 10 m2
Local integration around column T4
TAC = $25,000/yr
Potential Penalty = $35,000/yr
Heat-recovery match (T4 bottoms against T4 preheat) requires 39 m2
Cooler on T4 bottoms area = 36 m2
Make-up heater
TAC = $59,000/yr
Potential Penalty = $2,000/yr
Area = 4 m2
Heat recovery between T3 feed and T1 bottoms
TAC = $28,000/yr
Potential Penalty = $12,000/yr
Heat-recovery exchanger area = 41 m2
Cooler on T3 feed area = 55 m2
Heat recovery between raffinate and T1 tops
TAC = $55,000/yr
Potential Penalty = $16,000/yr
Heat-recovery exchanger area = 241 m2
Cooler on raffinate area = 77 m2
Heat recovery between T2 bottoms and T1 tops
TAC = $92,000/yr
Heat-recovery exchanger area = 75 m2
Heater on T1 tops area = 5 m2
Cooler on T2 bottoms area = 36 m2
Final TAC = $676,000/yr

(Stream 10) from the problem and develop a separate subnetwork for this
system. It consists of a heater having
114 m2 of surface, a heat-recovery
exchanger of 714 m2, and a cooler of
10 m2.
Decomposition analysis of the remaining eight streams is now undertaken. The minimum TAC for the
subsystem is $194,000/yr. The analysis indicates that the problem can
again be broken down into two sections, the first containing Streams 1 to
6 (TAC = $204,000/yr) and the second containing just Streams 7 and 8
(TAC = $25,000/yr). Here, too, the
second group is an actual heat-recovery match. The TAC for the decomposed problem is $229,000/yr, with a
potential cost penalty of $35,000/yr
(6% of the theoretical minimum).
Separate the T4 preheat (Stream 7)
and the T4 bottoms (Stream 8) from
the problem and integrate them separately. The optimum temperature difference for this subsystem is 41C
and the associated TAC is $25,000/yr.
The network consists of two units, a
heat-recovery unit of 39 m2 and a
cooler of 36 m2.
Decomposition analysis of the remaining six streams is now undertaken. First consider using utility to heat
the make-up stream (Stream 1). The
TAC for the six-stream problem is, as
noted above, $204,000/yr. For the decomposed problem, the TAC is
$206,000/yr.
Heat the make-up stream using hot
utility. This requires a heater of 4 m2
and has a TAC of $59,000/yr.
Decomposition analysis is applied
to the remaining five streams. All
possible decompostions are found to
involve significant penalty. Thus, attention should now be directed to applying decomposition analysis on a
thermal basis.

Stage 5 Decomposition
on a thermal basis
The five streams remaining
listed in Table 2. These streams
now ordered in accordance with
hottest target temperatures for

are
are
the
the

Figure 8.
30

48

80

48

140

Raffinate

828
30

Structure derived
using the pinch
design method.

T2 Bottoms

252
T1 Tops

38

80

1472
264
1288

cold streams and hottest supply temperatures for the hot streams. Temperature spans and stream heat loads
provide secondary guidance with regard to the ordering.
Stream 1 should be T1 tops (the
cold stream with the hottest target
temperature). The next stream should
be the hot stream with the highest
supply temperature, which is T2 bottoms. This hot stream has insufficient
load to fully handle the cold stream.
So, the next stream should also be a
hot stream. Of the remaining hot
streams, the raffinate has the highest
supply temperature. The remaining
two streams can be placed in either
order. We choose to place the hot
stream first. So, the ranking is now:
Stream 1 = T1 tops;
Stream 2 = T2 bottoms;
Stream 3 = raffinate;
Stream 4 = T3 feed; and
Stream 5 = T1 bottoms.
Decomposition analysis is then applied to this order. The TAC of the
overall subsystem remains at
$147,000/yr. However, there is now a
suitable decomposition involving a
system containing Streams 1 to 3 and
one containing an actual match between Streams 4 and 5. The actual
cost of the match between the T3 feed
and the T1 bottoms is found to be
$28,000/yr. The minimum theoretical
cost for the three-stream system is
$131,000/yr. The combined costs are
$159,000/yr, which has a potential
cost penalty of $12,000/yr.
Match Streams T3 feed and T1
bottoms. A point target exists at the
identified optimum approach of 16C.

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

Figure 9. Alternative serial structure.

The network contains two units, a

heat-recovery exchanger of 41 m2 and
a cooler of 55 m2.
Now just three streams are left.
The optimum temperature approach
for the subnetwork involving these
streams is 10C and the TAC is
$131,000/yr. A point target at this
temperature approach indicates that
the pinch occurs at a cold stream temperature of 38C and that 264 kW of
hot utility and 1,080 kW of cold utility are required. Application of the
pinch design method to this problem
would result in a network containing
a stream split as shown in Figure 8.
An alternative serial arrangement is
shown in Figure 9. First, the raffinate
(the hot stream with the supply temperature closest to the pinch) is
matched against T1 tops. Range targeting indicates an optimum temperature approach of 10C. The area required for the raffinate cooler is 77 m2,
and the area of the heat-recovery exchanger contacting the raffinate with
T1 tops is 241 m2. The remaining duty
on the T1 tops stream is 1,552 kW.
Thus, the temperature of the stream
leaving the heat-recovery unit is: Tt

H E AT T R A N S F E R

1,552/CP = 80 1,552/72 = 58.4C.

The T1 tops stream is now divided
into two parts: the first element involves heating the stream from 38C
to 58.4C, the second heating from
58.4C to the final target temperature
of 80C. Matching the raffinate with
the first element results in a subsystem of with a TAC of $55,000/yr. Optimizing the match between the remaining part of the T1 tops stream
and the T2 bottoms stream indicates
an optimum temperature approach of
10C and a TAC of $92,000/yr. The
T2 tops cooler has an area of 36 m2,
the remaining heater has an area of 5
m2, and the heat-recovery exchanger
has an area of 75 m2.
The overall TAC of the serial arrangement is $147,000/yr. This is
close to the minimum value of
$131,000/yr, which would require the
stream-split arrangement. The serial
design is accepted.
The final design is summarized in
Table 3. The final TAC is $676,000/yr,
which is 16.4% higher than the theoretical minimum cost. The differences
are accounted for as follows: integration around the reactor: 5.2%; integration around T4: 6.0%; make-up
heater: 0.3%; T3/T1 recovery: 2.1%;
the serial subnetwork: 2.8%; and the
application of local integration:
11.5%. The piping costs associated
with the remainder of the scheme
have not been determined.

Implications for process

integration software
The procedures described here
have major benefits. They allow the
designer to produce energy-efficient
network structures that are simple.
They control piping costs and avoid
the hazards associated with over-integration. They simplify the network
design process. They avoid the need
for sophisticated network-optimization procedures and can be employed
using any process-integration targeting program. This has implications
for software vendors and purchasers.
Currently, process integration software falls into two categories. There

are expensive programs that provide

targeting, network design, network
analysis, and network optimization.

Literature Cited
1. Umeda, T., et al., Heat Exchange System Synthesis, Chem. Eng. Progress, 74
(7), pp. 7076 (July 1978).
2. Umeda, T., T. Harada, and K. Shiroko,
A Thermodynamic Approach to the
Synthesis of Heat Integration Systems in
Chemical Processes, Computers and
Chemical Engineering, 3, pp. 273282
(1979).
3. Umeda, T., K. Niida, and K. Shiroko,
A Thermodynamic Approach to to Heat
Integration in Distillation Systems,
AIChE Journal, 25 (3), pp. 423429
(May 1979).
4. Itoh, J., K. Shiroko K., and T. Umeda,
Extensive Use of the T-Q Diagram to
Heat Integrated System Synthesis, presented at the International Symposium
on Process Systems Engineering, Kyoto,
1982, and published in Computers and
Chemical Engineering, 10, pp. 5966
(1986).
5. Amidpour, M., and G. T. Polley, Application of Problem Decomposition in
Process Integration, Transactions of
IChemE, 75A, pp. 5363 (1997).
6. Ahmad, S., and B. Linnhoff, Supertargeting: Different Process Structures for
Different Economics, Journal of Energy Resources Technology, 111, pp.
131136 (1989).
7. ESDU International, PLC, Costing of
Shell-and-Tube
Heat
Exchangers,
ESDU, London (1999).
8. Amidpour, M., and G. T. Polley, Dealing With Piping Constraints in Heat Exchanger Network Synthesis, IChemE
Symposium on Process Integration and
Fluid Separations, Manchester, U.K.
(June 1994).
9. Ponton J. W., and R. A. B. Donaldson,
A Fast Method for the Synthesis of Optimal Heat Exchanger Networks, Chem.
Eng. Sci., 29, pp. 23752377 (1974).
10. Ahmad S., B. Linnhoff, and R. Smith,
Design of Multipass Heat Exchangers:
An Alternative Approach, Journal of
Heat Transfer, 110, pp 304309 (May
1990).
11. Saunders, E. A. D., Heat Exchangers:
Selection, Design and Construction,
Longman Group, London, copublished
in the U.S. by John Wiley & Sons, New
York (1988).

CHEMICAL ENGINEERING PROGRESS DECEMBER 1999

There are much-less-expensive programs that concentrate on targeting

calculations. The techniques presented here favor these less-expensive
programs. [See the CEP Online Software Directory at http://www.aiche.
org/software/softwareindex.htm for
information on specfic software packCEP
ages. Editor]

G. T. POLLEY (Phone: +44-1229-585-330;

Fax: +44-1229-585-708; E-mail:
gtpolley@compuserve.com) is now retired
from the Univ. of Manchester Institute of
Science and Technology, Manchester, U.K.,
and is consolidating and publishing his
research. Following PhD research into
condensation heat transfer, he joined the
U.K.s National Engineering Laboratory in
1975, and was subsequently responsible for
research into boiling-heat-transfer and heatrecovery system design. Since 1978, he has
been a member of the International Energy
Agency Working Party on Heat Exchanger
Technology. In 1985, joined the chemical
engineering department at the Univ. of
Manchester Institute of Science and
Technology, as the director of the Centre for
Process Integration. His own research
activities centered around the development of
process and equipment design
methodologies. In 1990 was awarded the
Moulton Medal by the Institution of Chemical
Engineers in recognition of work on oil
refinery revamping, and in 1992 was awarded
the Ackrill Trophy by the U.K. Heat Transfer
Society for work on the applications of heattransfer enhancement. He holds BTech, MSc,
and PhD degrees in chemical engineering
from Loughborough Univ. of Technology, and
is a member of AIChE.
P. J. HEGGS is a professor and head of the
chemical engineering department at the Univ.
of Manchester Institute of Science and
Technology, Manchester, U.K. (Phone: +440161-200-4370; Fax: +44-0161-200-4399; E-mail:
p.j.heggs@umist.ac.uk). Previously, he held
positions as a professor at the Univ. of
Bradford, a lecturer at Leeds Univ., and a
senior engineer at Union Carbide Corp.s
Technical Center in South Charleston, WV. He
has been involved in heat-transfer research
for over 35 years, and his other interests
include adsorption/desorption, reaction
engineering, and mathematical modeling of
engineering equipment and processes. He
holds a BSc and PhD in chemical engineering
from Leeds Univ., and he is a Fellow of the
Institution of Chemical Engineers, a Fellow of
the Royal Academy of Engineering, and a
Chartered Engineer.