Rules for paths construction for HENs debottlenecking

Rules for paths construction for HENs debottlenecking

P.S. Varbanova, J. KlemesÏ b,*

aInstitute of Chemical Engineering, Bulgarian Academy of Sciences, BulgariabDepartment of Process Integration, UMIST, PO Box 88, Manchester, M60 1QD, UK

Abstract

This paper is based on the heat exchanger network retro®t techniques, developed by Tjoe andLinnho� and extended by Asante and Zhu. It considers, under the Network Pinch framework, twoimportant cases Ð the Retro®t Initialisation and Topology Modi®cation when the direct application ofthe classic Network Pinch concept and rules is not possible. With the help of a system of simpleheuristics, these limitations are overcome which extends the application range of the Network Pinchframework. 7 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Heat exchanger networks; Retro®t; Heuristics; Paths; Retro®t initialisation

1. Introduction

The heat exchanger network retro®t methodologies of Tjoe and Linnho� [8] and Asante andZhu [1,2] are based on the well established principles of Pinch Analysis [4,5] and its relatedconcepts [3,7], developed at the DPI UMIST. Tjoe and Linnho� provided the basics of settingretro®t targets and a rigorous-pinch-division retro®t strategy. First, from the existing network,which usually consumes extra energy, a MER topology is produced. It is then evolved usingpath and loop trade-o�s to reach some economically reasonable solution. More recentlyAsante and Zhu introduced the concept of the Network Pinch and a strategy and procedurefor a stepwise iterative topology improvement. The Network Pinch retro®t procedure uses

Applied Thermal Engineering 20 (2000) 1409±1420

1359-4311/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.PII: S1359-4311(00)00015-6

www.elsevier.com/locate/apthermeng

* Corresponding author.E-mail address: [email protected] (J. KlemesÏ ).

mathematical programming at its diagnostic stage to ®nd the most promising topologymodi®cation at each step.

1.1. Why ``paths construction''?

The present work is an extension of the Network Pinch methodology for cases when aNetwork Pinch cannot be identi®ed. It also facilitates the choice of topology modi®cation inthe Network Pinch framework. The following structure modi®cations are considered: (i) matchrelocation (including one-side), (ii) match addition, (iii) match removal, (iv) splits addition, (v)splits removal. As the Network Pinch retro®t strategy relies mainly on paths and loops trade-o�s to identify the bottlenecks, this is, consequently, the main objective of the present study.

1.2. Analysis

1.2.1. A special case Ð retro®t initialisationIn certain heat exchange systems, a Network Pinch cannot be found, as there is no path for

utilities-recovery trade-o�, despite the poor level of their heat recovery. An example of such asystem is given in Fig. 1. As it can be seen, the single process to process match does notproduce a load shift path.In such cases, there is a need for a guideline as to where to start and where to direct the

retro®t actions.

1.2.2. Modi®cations selectionIn order to select the most promising modi®cation in Network Pinch methodology, MILP

models are endorsed. This is a sort of combinatorial optimisation, where the topology changealternatives are compared, based on the rule that a promising modi®cation should move heat

Nomenclature

T(high) higher temperature of a cold stream (Table 1)T(low) lower temperature of a cold stream (Table 1)Cp heat capacity ¯owrate [kW/8C]DH enthalpy change [kW]A, B, C heat transfer area cost law coe�cientsTT(max) the maximum temperature, which the considered cold stream can reach, based

on the supply temperature (TS) of a given hot streamLoad(max) greatest possible load, based on TT(max)DTmin minimum allowed temperature di�erence

the symbol `` . . . '' stands for a number. This is a designation for a heatexchanger, existing before the retro®t.The symbol `` . . . '' stands for a number. This is a designation for a heatexchanger, added during the retro®t

P.S. Varbanov, J. KlemesÏ / Applied Thermal Engineering 20 (2000) 1409±14201410

from below to above the Network Pinch. There are cases, however, when a Network Pinch canbe identi®ed, but no modi®cation using this classic heuristic can be found. Consequently, thereis a scope for introducing some guidelines for obtaining an ordered set of topologymodi®cations, based on a system of additional heuristics which complement the classicNetwork Pinch search mechanism.

2. Paper goals

The analysis gives the directions for improvement of the existing methodologies. These are:

. To handle, by Retro®t Initialisation, the cases when no Network Pinch can be established.

. To facilitate the formation of an ordered set of topology alteration options thus enabling theprogress of the topology retro®t process.

3. Methodology

As it is mentioned above, for improvement of the network structure, path and loopinteractions are used. When such paths are not available or the potential of the existing pathsis exhausted, then a necessity for creation of new paths is established. Below is given theheuristics system, used for paths generation. The rules are ordered by their relevance andgrouped by the context they are best illustrated.

3.1. General topology path-rules

3.1.1. Retro®t initialisationThe ®rst two rules give the necessary conditions for the existence of complete trade-o� paths,

Fig. 1. Example 1 Ð a HEN with poor heat recovery, without a Network Pinch.

P.S. Varbanov, J. KlemesÏ / Applied Thermal Engineering 20 (2000) 1409±1420 1411

usable for heat recovery improvement. As they represent basic concepts, they are bestillustrated by the Retro®t Initialisation case.

(G1) Start and end of the paths. The paths has to connect:(a) Hot with cold utilities (Fig. 2). In this case, according to the +/ÿ principle, anyincrease of recovery through the path produces a simultaneous decrease in hot and coldutility demands.(b) Hot with hot or cold with cold utilities with di�erent prices. In this case the generalrule is that for a given stream expensive utilities require less heat-exchange area and vice-versa. In this case it is possible to establish some area distribution trade-o�s.

(G2) Middle points. The paths have to include at least one recovery (process to process)match. The trade-o� options, produced by the new paths, in general, involve an increase inthe number and complexity of the interdependencies among the network variables. This, inturn, results in more complex parameter optimisation and more complex HEN control. Inorder to minimise the above e�ect, the following rule is introduced:(G3) The path length should be as short as possible. Thus, the path a�ects fewer HENvariables, and is easier to optimise and control.

3.2. Retro®t-speci®c heuristics

(R1) In the given context of options for matching a certain stream, a match, which can beimplemented with minimum additional area (ideally zero) is preferable.In the course of the parameter trade-o�, it is possible to reduce the load of a certain utility

match to zero resulting in match elimination. In this case, the exchanger of the eliminatedmatch may be re-used for implementation of a new or extended recovery match in the trade-o�

Fig. 2. Example 1 Ð retro®t initialisation utility-recovery trade-o� paths.


path to reduce the additional area involved. Such a type of topology modi®cation is preferableto modi®cations that reduce the loads of utility matches without eliminating them.

3.3. Heat-exchange process heuristics

These rules are intended to help the engineer to obtain an ordered set of alternative topologymodi®cations. Furthermore, based on these rules, it is possible, by selecting several topologymodi®cations at once, to construct a modi®cation plan looking several steps ahead.

(HE1) Having a stream to match, ®rstly temperature-feasible candidate streams are identi®ed.This rule substantially reduces the amount of information, which has to be processedfurther. Consider hot streams 7 and 8 from Example 1. The potential match cold streamsare those from 1 to 5 (see Fig. 1). However, only the streams 1 and 4 are feasible matcheswith 7 and 8 (see Table 1). Therefore, any further considerations will only involve the abovetwo cold streams instead of the entire set of cold streams.(HE2) Streams with closer heating/cooling demands are preferable for matching. This ruleshould reduce the number of excessive small-load matches. The heat demands of both coldstreams in Table 1 are closer to the heat available from stream 8. Consequently we arerequired to match stream 8 with stream 1 or stream 4.(HE3) Having a cold stream to heat, among the available hot streams, a match with the higher-temperature stream is more desirable. Having a hot stream to cool Ð the heuristic issymmetrical. For an example, this rule suggests again the choice of stream 8 as a matchcandidate for the only free cold stream 1. A more enhanced form of rule (HE3) for the casewhen we have many hot and potentially matching cold streams is: ``Sort hot streams:descending, ®rst according to their supply and then Ð target temperatures; cold streams:descending, ®rst according to their target and then Ð supply temperatures''. This shouldguide the design to equable driving force distribution and hence to the minimal additionalarea. An example of almost fully sequentially ordered system of hot and cold streams isshown below (Fig. 3).

The above three rules constitute the basis for a systematic retro®t initialisation. Based onthese, in Example 1, a new exchanger (E1) is added. The best duty for this heat exchanger is itsmaximum of 200.26 kW. The Network Pinch can be readily identi®ed at the hot side of thenewly placed match. The retro®t process can now continue according to the Network Pinchmethodology through the path H5-4-E2-C7 as is shown in Fig. 4 below.

Table 1Example 1 Ð cold streams, feasible for a match with 7 and 8

Stream T (high) T (low) CP DH

1 (part) 68.00 58.50 21.08 200.26 Free

4 (part) 68.00 45.00 13.73 315.79 Used in match (4)


If, however, the number of matching possibilities is larger, then the following additionalrules may be applied:

(HE4) Matches between streams with closer heat capacity ¯owrates are preferable. It isbecause such a match will feature better logarithmic mean temperature di�erence and lowerarea requirement at a given load.(HE5) Matches between streams with closer ®lm-transfer coe�cients are preferable. Theapplication of this rule should also result in lower area requirement through raising thee�ectiveness of the heat transfer process.

Fig. 3. Equable driving force distribution for the process from Example 2.

Fig. 4. The retro®t process after the initialisation.


(HE6) If there are certain streams that require heat recovery and their CPs are (substantially)smaller than the CP of the potential match-stream, splitting is required. This will makefeasible matches at greater loads.

3.4. Topology modi®cations selection

The classic Network Pinch methodology suggests a strategy for topology modi®cationassessment, which is based on the following approach: (a) Higher recovery potential; (b)Smaller additional area. The here presented heuristic selection of topology changes is alsobased on this philosophy. The matches' loads directly assess the higher recovery, whichconstitutes the ®rst part of the strategy. The second, additional area minimisation, is theobjective of rules HE2 to HE5.An additional rule involving the design economics is:

(HE7) At the overall network level, the achievement of maximum heat recovery with theminimum number of additional matches is desirable.

A match, which will cause load splitting into extremely small exchangers is not desirable.This rule results from the consideration of the area cost law, which can be written in thefollowing form:

Cost � A� B � AreaC �1�It is easily seen that the coe�cient ``A'' would have a smaller impact when fewer number oflarger-size exchangers are used.Now consider Example 2, a sun¯ower oil production plant, ®rst analysed by Nenov and

Kimenov [6]. The current HEN with DTmin of 68C is given in Fig. 5. Pinch Analysis providesthe following data: pinch location at 26=208C; total utility heating of 316.44 kW; total utilitycooling of 21.84 kW, targeted sum of the total heat recovery 986.36 kW.After pushing the load of match (1), the Network Pinch is found at the cold side of

exchanger (1) (Fig. 6). The load of the only recovery match (1) of 721 kW is equal to 83.64%of heat recovery with respect to the MER target. Although, the Network Pinch is identi®ed, itcan not be overcome with the help of the classic rule for heat transfer from below to above theNetwork Pinch. Consequently the rules HE1 and HE3 are applied ®rst. A list of streams,feasible for matching with stream 10, ordered on TT(max), TS and Load(max) are obtained(Table 2), where TT(max) is the maximum temperature the considered cold stream can reach,based on the supply temperature (TS) of stream 10. Load(max) is the greatest possible load,based on TT(max).Indeed, the best ®t to the heat remaining on stream 10 (140.74 kW) is sought. As can

be seen from the check-list, the best ®t comprises two matches (with cold streams 4 and2) and leaves a residual duty on hot stream 10 of 2.44 kW (Fig. 7). The new topology,with loads maximised on the new recovery matches is shown in Fig. 5. The amount ofheat recovering after the placement of (E2) is 79.30%, and after the placement of (E3) is87.21%. The available heat on stream 10 has been reduced at 2.44 kW. Let us now look


Fig. 5. Example 2 Ð the initial network.

Fig. 6. Example 2 Ð the Network Pinch for the existing HEN.


at hot stream 11 Fig. 8. First the cold streams have been identi®ed as possible matcheswith 11 (Table 3).Streams 5 and 1 are chosen as best candidates, with loads of matches 12.16 kW from the

total 28.56 kW for 5 and 111.60 kW Ð the whole demand of stream 1. The modi®ed topologyis shown below:The two retro®t steps passed show one of the great advantages of the presented topology

modi®cation approach. Here, at the second step, a bene®cial thermal pro®le of streams 5 and 1

Table 2Example 2 Ð cold streams feasible for matching with stream 10

Sort on TT(max), TS, Load descending Selection

Stream TS (8C) TT (8C) TT (max) (8C) Load (kw) Cp (kw/8C) Selected Load (kW)

6 75.00 90.00 85.00 40.40 4.04 0.00

4 70.00 100.00 85.00 60.30 4.02 Z 60.305 20.00 90.00 85.00 54.60 0.84 0.002 50.00 70.00 70.00 78.00 3.90 Z 78.00

3 20.00 70.00 70.00 4.00 0.08 0.001 20.00 50.00 50.00 111.60 3.72 0.008 85.00 90.00 85.00 0.00 4.04

Total selected load 138.30

Complement needed 2.44

Fig. 7. Example 2 Ð recovering heat from stream 10.


is identi®ed. As a result, it is chosen to add two matches at once. Such opportunities cansubstantially simplify the modi®cations search process.At this stage the heat recovery has increased to 99.75%. Is it now possible to recover the

rest of the heat from (C1), and at what cost? The answers are: ``yes'' and at zero cost. There isa possibility to re-use the existing cooler on C1, for recovery purposes. Stream 10 supplies itsresidual heat at 60.548C. From the cold streams, which still need process heating, feasible forthe heat exchange with stream 10 are streams 3 and 5. As it can be seen, both options arefeasible and with identical supply temperature. As exchanger (C1) is initially sized to handle asubstantially greater amount of heat, it makes no di�erence what option we choose. Anyfurther choice is arbitrary and may be based, say, on layout considerations. However, wechoose to match cold stream 5 for this example. Thus, the ®nal topology features 100% heatrecovery (Fig. 9).

Fig. 8. Example 2 Ð recovering heat from hot stream 11.

Table 3Example 2 Ð cold streams, feasible for matching with hot stream 11

Sort on TT (max), TS, Load descending ActualStream TS (8C) TT (8C) TT (max) (8C) Load (kw) Cp (kw/8C) Selected Load (kW)

5 20.00 90.00 54.00 28.56 0.84 Z 12.163 20.00 70.00 54.00 2.72 0.08 0.001 20.00 50.00 50.00 111.60 3.72 Z 111.60

Total 123.76

Complement needed 0.00


4. Conclusions

The proposed system for heuristic topology modi®cation is a complement to the NetworkPinch methodology for heat exchanger network retro®t. It considers two important cases inwhich the classic Network Pinch methodology is not directly applicable.The ®rst is the case of Retro®t Initialisation when a Network Pinch cannot be identi®ed. The

application of this new approach provides the opportunity to exploit the power of the NetworkPinch concept and framework for a more broad range of heat exchanger networks.The second case is the enhancement of topology modi®cations selection in which heat cannot

be transferred from below to above the Network Pinch. The presented systematic approach,built on a system of simple heuristic rules, obtains an ordered set of topology alterationalternatives, and in some cases identi®es a topology modi®cations sequence in one step, whichmay substantially simplify and speed-up the modi®cation procedure.

References

[1] N.D.K. Asante, X.X. Zhu, An automated approach for heat exchanger network retro®t featuring minimaltopology modi®cations, Comp. Chem. Engng 20 (Supplement) (1996) S7±S12.

[2] N.D.K. Asante, X.X. Zhu, An automated and interactive approach for heat exchanger network retro®t,

Transactions of IChemE, 75 (Part A) March (1997).[3] B. Linnho�, S. Ahmad, Cost optimum heat exchanger networks. Part I: Minimum energy and capital using

simple models for capital cost, Computers Chem. Engng 14 (1990) 729±750.

Fig. 9. Example 2 Ð the MER topology.


[4] B. Linnho�, E. Hindmarsh, The pinch design method for heat exchanger networks, Chem Eng Sci 38 (1983)745±763.

[5] B. Linnho�, et al., User Guide on Process Integration for the E�cient Use of Energy, Pergamon Press, NewYork, 1982.

[6] N. Nenov, G. Kimenov, Energy conservation through process integration in sun¯ower oil production, Scienti®c

Works of HIFFI Plovdiv (in Bulgarian), XLII (1997).[7] R. Smith, in: Chemical Process Design, McGraw-Hill, New York, 1995, p. 459.[8] T.N. Tjoe, B. Linnho�. Pinch technology retro®t: setting targets for an existing plant, Chemical Engineering,

April 28 (1986).


Rules for paths construction for HENs debottlenecking

Documents