Computers them. Engng Vol. 19, SuppI.,pp. S143-S148.

1995
Copyright 81995 Elsevier Science Ltd
Pergamon 0098-1354(95)00166-2 Printed in Great Britain. All rights reserved
0098-1354/95 $9.50 + 0.00

THE PLACEMENT OF TWO-STREAM AND MULTI-STREAM HEAT-EXCHANGERS

IN AN EXISTING NETWORK THROUGH PATH ANALYSIS

JOS L.B. VAN REISEN*, JOHAN GRIEVINK*,

GRAHAM T. POLLEY** AND PETER J.T. VERHEIJEN*

*Department of Chemical Engineering, Delft University of Technology

Julianalaan 136, 2628 BL Delft, The Netherlands
fax ++31 - 15 - 784452, E-mail [email protected]

**Department of Chemical Engineering

University of Manchester Institution of Science and Technology
P.O. Box 88, Manchester M60 lQD, United Kingdom

ABSTRACT

A prescreening and decomposition method is presented to analyse heat exchanger networks for
retrofitting. The method, called Path Analysis, selects and analyses fractions from the existing network,
either by heuristics or by an algorithm. By comparison of all fractions, the critical parts of the network
that should be adapted can be identified. The adaptations can be done independent of the remaining
network. Thus Path Analysis enables a considerable reduction of the effort in retrofit design. Mean-
while the simplest network adaptations are favoured.
Path Analysis is applied to several cases. The results for an aromatics case are presented. Using the
right software tools, the engineering effort can be reduced considerably, compared with existing
methods. Solutions tend to be less complex, while the profitability is sometimes higher than was
expected from global analysis. With Path Analysis the retrofit design using new multi-stream heat
exchangers proved to be straightforward.

KEYWORDS

Heat Exchanger Networks; Retrofit; Targeting; Compact Multi-Stream Heat Exchangers

INTRODUCTION

The major part of the available heat integration literature addresses methods developed for grassroots
designs (Linnhoff, 1993). Although these methods are valuable for retrofit designs, specific methods
should be used to properly find and evaluate possible adaptations to the existing equipment. Full
retrofit analysis and design methods have been reported by Tjoe (1986), Carlson et al. (1993) and Yee
and Grossmann (1991). The former two report on heuristic evolutionary methods, the latter on math-
ematical programming. All methods perform both the retrofit analysis and design on the complete heat
exchanger network that exists. Irrespective of the used method the design of even a near optimal
network is complex task. This complexity increases rapidly with the size of the retrofit problem. Tjoe
(1986) presented retrofit targets to guide the designer towards at least near optimal networks. Carlson
et al. (1993), however, recognized that these targets are of limited use, as the costs of different new
matches may differ substantially. The designer may thus overlook more cost effective retrofit designs.

A possible way to reduce the complexity is some prescreening procedure that not only estimates an
economic heat recovery level, like in previous methods, but also identifies the critical parts of the
existing network that should best be adapted. The procedure should thus reduce the design problem
without the loss of relevant energy saving potential. Daichendt and Grossmann (1994) have published
a mathematical procedure to do a similar screening of grassroots cases. Hui and Ahmad (I 994) have
published a method to handle zones of integrity, that may also be used. In a study (Van Reisen, 1993)

s143
s144 European Symposium on Computer Aided Process Engineering-5

on the use of compact multi-stream heat exchangers in retrofit design a new retrofit analysis tool has
been developed. This tool, called Path Analysis, generates a series of subproblems from the total retro-
fit problem that have high economic potentials. It enables a designer to trade of before design energy
conservation, area addition and expected complexity of both the design task and the design itself.

PATH ANALYSIS

Concents

Path Analysis is a simple decomposition and prescreening procedure that evaluates the economic
potential of parts of the existing heat exchanger network, called subnetworks, while the remaining part
of the network is unchanged. The evaluation of each subnetwork is done by existing retrofit analysis
procedures, So, targeting is performed for energy savings and investments in each subnetwork. The
targets of each subnetwork can be compared with each other and with the total network. From this
comparison the most promising set of subnetworks is selected and a retrofit design is made, using one
of the existing procedures.

The subnetworks can be generated either by the designer based on his knowledge of the process or by
an algorithm to allow rigorous analysis of all opportunities. Two rules should always be taken into
account. First, all subnetworks should be heat-balanced. This means that heat exchangers and the
stream sections covered by them should be selected from the existing network rather than streams with
a source and target temperature. If this rule is not obeyed the subnetworks may not be adapted
independent from the total network. Second, at least one heater and one cooler should be included in
each subnetwork. This is a minimal requirement for an opportunity for energy conservation. In more
general terms the subnetwork should consist of a ‘flexible’ heat sink, including one or more heaters,
and a ‘flexible’ heat source, including one or more coolers. To both the sink and the source process-to-
process heat exchangers can be added.

Another requirement for energy conservation is a connection between the heat sink and the heat
source. Such a connection, called a path (Linnhoff and Hindmarsh, 1983), can be used to reduce the
loads on both the sink and the source. In order to do this, energy is transferred from the source to the
sink by process-to-process heat exchange. To save energy, area must be added to existing exchangers
in an existing path or to new matches, that create new paths. Adding area to existing exchangers is
preferred as it has the lowest impact on the existing structure and it is relatively cheap. When subnet-
works are build from fractions of existing paths, the number of new matches needed is as small as
possible. This guideline can help to identify promising subnetworks from the total network. It enables
the evaluation of the expected complexity of the design task and the design result at the same time the
savings and investments targets are available.

Procedure

The steps for retrofit design with Path Analysis are presented in Fig. 1. After data extraction, the
saving on investment relationship for the total network are calculated. Any of the existing retrofit or
eventually grassroots targeting methods can be used. Also the thermodynamic minimum energy input
is determined from the energy target at the minimal approach temperature of 0°C. Both the savings on
investment relationship and the minimum energy are used as a reference for further analysis.

The first stage of the Path Analysis is the identification of the subnetworks in which energy savings
can be established followed by an analysis of these subnetworks similar to the analysis of the total
network. In the previous paragraph the two requirements for such subnetworks have been identified. If
the problem is sufficiently small or if appropriate software is available, all possible subnetworks from
the simplest paths to the total network, can be evaluated. The number of possibilities increases rapidly
with the number of streams, the number of existing matches and the number of heaters and coolers.
 European Symposium on Computer Aided Process Engineering-5 s145

, Extract data existing network

Determine
overall savings/investment

Select subnets with comparable/better than overall savings/investment

I

Rank subnets on profitability/complexity/operability ,

Create new design for (next) best subnet

I
Evaluate design
I versus target

New Retrofit Design

Fig. 1. Schematic representation of the Path Analysis

The second stage is a comparison of all analysed subnetworks. This comparison should not only
include the savings and investments, but also the complexity and the practical problems or
opportunities of the considered subnetworks. The complexity is primarily determined by the number of
streams and the number of exchangers that exist. The size of those exchangers may also be a
complicating factor as large areas may be difficult to reuse. Several practical aspects also have
influence on the complexity. Streams can be far apart in physical space while others are connected or
related according to the flowsheet or plotplans. The subnetworks may differ in the opportunities with
respect to physical space and existing civil works to put in new exchangers or in problems related to
controllability, flexibility and safety. All these practical aspects greatly influence the total costs of the
retrofit project. The currently known cost functions cannot take into account these aspects and thus a
somewhat subjective and only qualitative trade off must be done. The Path Analysis shows the energy
and area implications of certain choices.
The last stage of the Path Analysis is the creation of a retrofit design for the chosen subnetwork and a
comparison of the results with the targets. All streams and matches outside the subnetwork are
unmodified. The design phase can thus considerably be simplified. Generally the smaller the project
the smaller the chosen subnetwork will be. Thus the engineering effort can be adapted to the proposed
scope of the project. The design can be done by one of the methods mentioned in the introduction.

EXAMPLE

The path approach will be illustrated by the aromatics case, discussed by Tjoe (1986). The problem is
summarised in Fig. 2. The used cost data is similar: hot utility 0.177 Dfl/W/yr cold utility 0.02
Dfl/W/yr. The area costs per shell in Dfl were 24*103+ 1.9*103 (shell area) 0 *83 . To analyse the
network the incremental area efficiency (a)targeting procedure has been used.
S146 European Symposium on Computer Aided Process Engineering-S

MCP h
kW/ ‘C kWlm’ T

100 0.8

160 0.5

60 2

200 0.4

100 5

70 1

175 0.5

60 ’ 60 0.2

140 * 200 0.8

22.4 9.6 \ 1222

Fig. 2. Data of the aromatics case from Tjoe (1986). Q (kW) \ A (m2>

Investment
(Millions)
Fig. 3 Savings on investment plots from incremental a targeting of the aromatics case
subnetworks: A = Hl, El, ES, Cl; B = Hl, El, E4, C2; C = Hl, El, E3, C3;
D = H2, E2, E5, Cl ; E = H2, E2, E4, C2; F = H2, E2, E3, C3;
 European Symposium on Computer Aided Process Engineering-5 s147

The existing network contains 4018 m* process-to-process exchanger area and uses 27.1 MW hot
utility. The area target for the same utility use is 3445 m*, thus the existing area efficiency is 0.86.
The thermodynamic minimum hot utility use is 14.4 MW, allowing a maximum energy saving of
12.7 MW. The targeted savings on investment relationship is plotted in Fig. 3.

The next stage in the Path Analysis is identification of valid subnetworks. Rigorous analysis of the
system would require the evaluation of about 220 subnetworks. For illustration purposes only six of
them are selected, based on engineering insight in the problem. These subnetworks have been analysed
by the same targeting procedure as the total network (Fig. 3).

The savings on investment lines in Fig. 3 clearly show good potential for the retrofit of the total net-
work. Most of the subnetworks have little saving potential compared to the total network and relative-
ly poor rentability as well. The positive exception is subnetwork C, containing the exchangers Hl, El,
E3 and C3. It has a substantial saving potential of 4.7 MW and a rentability that is comparable to the
total network. Note that complexity is not taken into account in these plots. The curve of C is the only
one presented that does not reach infinite investments. This is because there is no process pinch point
in this subnetwork. The load on the heater becomes zero at a minimum approachttemperature of 14’C.
Such situations can be derived directly from the savings on investments plots. If a project is defined
with a payback time up to 2 years, subnetwork C will be preferred over the total network because of
its rentability and the low complexity. For a payback time abritrarily set to 1.5 years a saving of 3.6
MW can be obtained in the subnetwork, while only 3.1 MW can be obtained in the total network.

Additional area

New match

3513 4362

Fig. 4. Final design of the subnetwork with HI, El, E3 and C3 of the aromatics case

The design of the selected subnetwork C with the exchangers HI, El, E3 and C3 can easily be done
by Tjoe’s method or even by inspection. The result is presented in Fig. 4. The adaptations are clearly
very simple: one new match and two area additions to existing matches. The added amount of area
was 1467 m*. This can be reduced to 1300 m* when a split is allowed in stream 4. This is very near
the incremental 01area target of 1357 m2. The payback of the first solution is 1.6 year, of the latter
1.4 year. The first solution is identical to the solution found by Tjoe and Linnhoff. The design and
evolution effort was however much larger.

COMPACT MULTI-STREAM HEAT EXCHANGERS

Path Analysis was developed in a study to the retrofit design using compact multi-stream heat
exchangers. These exchangers can be a valuable option in some applications. The main advantages are
the potential reduced piping costs and the limited physical space needed for the equipment. Compact
S14l3 European Symposium on Computer Aided Process Engineering-5

surfaces allow a very small approach temperature and have low costs per unit area. However, mechani-
cal cleaning is not possible and therefore the streams should be relatively clean. To make full use of
the strengths of the multi-stream heat exchangers in retrofit designs, opportunities must be found
where concentration of new area is favourable within a limited set of streams that are not too far apart
in physical space. Path Analysis can very well be used to find such opportunities. The subnetwork
identified in the example shows an opportunity for either a three- or a four-stream heat exchanger, if
fouling is not a problem.

OPPORTUNITIES AND LIMITATIONS

The above presented Path Analysis lends itself naturally to automation of the procedures, as the
required computational effort is minimal compared to e.g. optimisation with mathematical program-
ming. Our present setup is to generate subnetworks with specially developed software and to do the
last design phase with a commercially available program (ADVENT in the case above). It has proven
to be easy to apply simple heuristic rules in order to reduce the number of subnetworks at the earliest
possible stage. However, this depends on the choice of economic criteria which varies from case to
case and we are exploring various possibilities. The ranking of the subnetworks (Fig. 1) is more
difficult to quantify and to implement in software. We have sofar made the savings on investment
plots available for visual inspection. On the one hand, it would be an advantage to have a ranking
criterium, on the other hand, the here presented approach allows the designer to add at the targeting
stage other relevant information. It also opens opportunities to include more details. For example,
different costs can be assigned to matches between different streams, as proposed by Carlson et al.
(1993). These different costs can depend on the distance between streams, corrosion and safety.
In conclusion, Path Analysis appears a simple approach to a complex problem. It is straightforward in
concept and flexible in implementation.

REFERENCES

Carlson, A., P. Franck and T. Bemtsson (1993). Design better heat exchanger network retrofits. Chem.
Eng. Prog., March 1993, 87 - 96.
Daichendt, M.M. and I.E. Grossmann (1994). Preliminary screening procedure for the MINLP synthes-
is of process systems - II. Heat exchanger networks. Comput. Chem. Eng., l8, 679 - 709.
Hui, C.W. and S. Ahmad (1994). Minimum cost heat recovery between separate plant regions,
Comput. Chem. Eng., 18,711 - 728.
Linnhoff, B. (1993). Pinch analysis - a state-of-the-art overview. Trans. IChemE., m, 503 - 522.
Linnhoff, B. and E. Hindmarsh (1983). The pinch design method for heat exchanger networks.
Chem. Eng. Sci., 38, 745 - 763.
Tjoe, T.N. (1986). Retrofit of heat exchanger networks. PhD. University of Manchester Institute of
Science and Technology
Van Reisen, J.L.B. (1993). Rational process design using multi-stream heat exchangers: integration in
retrofit. Final report of a design study for NOVEM. Delft University of Technology.
Yee, T.F. and E. Grossmann (1991). A Screening and Optimization Approach for the Retrofit of
Heat-Exchanger Networks. Ind. Eng. Chem. Res., 3, 146 - 162.

ACKNOWLEDGEMENT

This work has been funded through NOVEM, the Dutch Society for Energy and Environment, under
contract No. 33104/0131. We thank the Dutch Advanced Heat Exchanger Study Group and its
chairman Ir. A.R. Braun for the initiative and the support during the study.