You are on page 1of 8

Full Paper

Straightforward Solutions of Heat Exchanger Networks (HEN)


Problems
By Kurtreiner Schack*

Advantage is taken of an unused potential which lies in the evolutionary development method for heat exchanger networks
(HEN) proposed by Linnhoff and Flower to construct an algorithm which reduces recursively the number of units of a given
HEN problem. As a result, all networks are generated that fulfill the heating and cooling requirements of the problem and whose
number of units cannot be further reduced. Among the solutions obtained, those are selected which satisfy the condition of
minimum permissible temperature driving force. A C++ implementation of the algorithm has been applied to a number of
problems taken from the literature.

1 Introduction initial structure using an integer programming method


combined with evolutionary steps resulting in minimum cost
In the last thirty years, a large number of methods for the designs rather than solutions with a minimum number of units.
solution of HEN problems has been published. Among them, The paper is organized as follows: section 2 revises briefly
the evolutionary development (ED) method proposed by the methods of Linnhoff and Flower and gives an extension of
Linnhoff and Flower [1] offers special advantages. The main the ED method. Section 3 describes the evolutionary steps
characteristic of this method is the use of heat quantities, so- leading to the solutions of the HEN problems. Section 4 is a
called freedoms, instead of temperatures to describe a heat short note on the minimum number of units. Section 5 explains
exchanger. The authors give a set of rules to calculate how the the method following the evolution of a known example
values of the freedoms change when certain evolutionary (TC3). Section 6 shows how networks containing heat
steps are carried out. The method has been applied rarely exchangers with insufficient freedoms can be used and section
since its publication in 1978. It is a shortcoming of the method 7 summarizes results for a number of further examples taken
that it does not give explicit strategies to find appropriate from the literature.
evolutionary steps [2,3]. Calandranis and Stephanopoulos [4]
use the method to examine the influence of disturbances on
the operability of a network structure. 2 Methodical Basis
The concept of freedoms, however, allows very convenient
mathematical formulations. Therefore, it is worth while to 2.1 The Temperature Interval Method (TI)
extend the ED method in a way to overcome its shortcomings
and to develop a mathematical system which leads in a A HEN problem consists in the task of structuring the heat
straightforward way to the solutions of a HEN problem. transfer between a series of hot and cold streams. Each stream
It is beyond the scope of this paper to give a complete review is defined by three values:
of other published methods. Examples of alternative ap- ± the heat capacity flow rate: Cc for a cold stream and Ch for a
proaches include: hot stream, respectively
Floudas et al. [5] introduce a nonlinear programming ± the supply temperature Ts
formulation. This starts from a superstructure in which each ± the target temperature Tt
stream is split into a number of partial streams equal to the The table in which these values are listed is in the following
number of heat-exchanging matches in which the stream referred to as task table.
participates. In this way, solutions which need stream splitting The application of the ED method demands to develop,
can be found. from the values of the task table, an initial structure consisting
Trivedi et al. [6] proposed an evolutionary method using of heat exchangers, heaters and coolers. The best way to
systematic energy relaxation in the network. The authors do generate this structure is given by the TI method which was
not use freedoms, but they apply a loop-breaking procedure also proposed by Linnhoff and Flower [9]. Following this
similar to that described below. method, the heat transfer task is divided into so-called
Chakraborty and Ghosh [7] create networks using rando- subnetworks, each corresponding to a smaller temperature
mization techniques. The networks generated in this way can interval. Initially, each subnetwork (SN) is treated as a
serve as initial solutions for a nonlinear optimization program. separate problem. This problem is the creation of a structure
Ren et al. [8] start with a structure containing only one heat- of exchangers, heaters and coolers which satifies the heating
exchanging match for each stream. The authors develop the and cooling demands of the streams in the given temperature
±
interval. As far as possible, these demands are satisfied by heat
[*] Prof. Dr.-Ing. K. Schack, formerly SHG-Schack GmbH, Am Knickbusch 1, exchanger matches between hot and cold streams; the rest is
D-29362 Hohne, Germany. done by external heaters (H) or coolers (C). It is important to

Chem. Eng. Technol. 25 (2002) 11, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1107 $ 17.50+.50/0
0930-7516/02/1111-1107 $ 17.50+.50/0 1107
Full Paper

mention that within a subnetwork, a stream is allowed to pass The authors of [1] define the freedom Fas the larger of these
through only one exchanger. Any remaining utility (H or C) values.
must be placed downstream the exchanger. DTs is the smallest temperature difference between the two
For the heat exchange, a minimum allowable temperature streams in the countercurrent heat exchanger. With
difference DTmin must be chosen. It is assumed that the stream
flows in the heat exchangers are countercurrent. Finally, the r = Ch/Cc (5)
subnetworks are merged to form the complete network.
The number of heat exchangers in a subnetwork is equal to the position of the smallest difference is on the left-hand side
the number nss of cold or hot streams in the subnetwork, of the exchanger if r > 1, and on the right-hand side if r < 1.
whichever is smaller. The number nss is zero if there are only Here, left' is where the hot stream, and right' where the cold
hot or only cold streams. In those cases the total heating or stream enters the exchanger. In the grid representation of the
cooling duty is accomplished by the utilities. network used here, it is generally assumed that the higher
If a subnetwork consists of more than two streams, then temperatures occupy the left-hand side.
there is more than one way to connect the streams by In a valid solution of a HEN problem, all heat exchangers in
exchangers. The number of variants can be calculated by the the network must have zero or positive freedoms. Conse-
formula [10]1) quently, in every step towards a solution, the resulting changes
nss of the freedoms must be taken into account. Using the
nvar = ∏ (nsl − j + 1) (nsl > 0; nss > 0) (1) definitions Eqs. (3) to (5), the rules which have been given by
j =1 Linnhoff and Flower [1] to calculate the changes of the
freedoms can be written in simpler form [11]. The rules are:
where nsl is the larger number of the streams. When the 1) Shifting a cooler through an exchanger:
subnetworks are merged to form the complete network,
combinations of all variants of subnetworks are possible. DFc = ± Qc (6)
These different combinations are here referred to as main
variants. Their number is 2) Shifting a heater through an exchanger:
N sn
N var main = ∏ nvar(i ) (2) DFh = ± Qh (7)
i =1
The number of main variants can reach the magnitude of In both cases, the upper sign is valid when shifting in the
several powers of ten. However, the method described below direction of the flow, and the lower sign is valid when shifting
takes into account all possible combinations of matches against the flow. Eqs. (6) and (7) also cover the case of
between the streams in every stage of the evolution. We interchanging two exchangers.
conjecture that starting from any of the main variants the full 3) Merging two exchangers between the same streams:
range of solutions is reached. This conjecture is supported by a The freedom F(m) of the merged exchanger becomes
series of example calculations. But there is one important
premise: the chosen main variant must contain at least one F(m) = F(l) if r > 1 (8a)
match between every hot and every cold stream. This premise
can always be fulfilled by introducing exchangers with zero
heat load. F(m) = F(r) if r < 1 (8b)

2.2 ED Method F(m) = F(l) = F(r) if r = 1 (8c)

The ED method describes a heat exchanger by three values: where F(l) and F(r) denote the freedom of the initially left-
(i) its heat load, (ii) the identity of the two streams it hand side or right-hand side exchanger. The Eqs. (8) are
connects, (iii) its freedom. valid for both values Fc and Fh of the freedom.
For mathematical convenience we define two values of the 4) Balancing a heater (H) and a cooler (C) through an
freedom: exchanger (Fig. 1):
With DQ = min (Qc, Qh) the heat load of the exchanger
Fc = Cc (DTs ± DTmin ) (3) after the balancing is

Fh = Ch (DTs ± DTmin) (4)

±
1) List of symbols used at the end of the paper. Figure 1. Balancing a heater and a cooler through an exchanger.

1108 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1108 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 11
Full Paper

Q(after) = Q(befre) + DQ (9) of the ratio DFh/DQ. The table has been published previously
in a different form [11].
and its freedom is The corresponding changes of the freedom Fc follow from
Eqs. (3) and (5). The validity of Tab. 1 assumes that the
Fs(after) = Fs(befre) ± DQ (10) sequence of the exchangers is as given in Fig. 2, without other
exchangers, heaters or coolers between them. Using the rules 1
with Fs = min (Fc, Fh). It is assumed that heater and cooler and 2 of the ED method, one can always shift the units in the
are placed in the direction of the flow after the heat network until the required sequence is realized.
exchanger. Heaters and coolers can be included into the loop-breaking
process if one looks at the heating or cooling medium as an
additional stream. In this way, heaters or coolers can be
2.3 Extended ED Method viewed as heat exchangers with infinite freedom.

For full use of the ED method, the case of eliminating


redundant heat exchangers must be included. This procedure 3 Proposed Method of Network Development
is known in the literature as loop breaking'. A loop is given if
there are more heat-exchanging connections between the The method develops an initial network by eliminating heat
streams than necessary and so heat flows in a cycle. Su and exchangers and utilities until all networks are obtained that
Motard [3] define loops of different levels corresponding to fulfill the following conditions:
the number of streams involved. A first-level loop concerns 1) They comply with the heating and cooling requirements of
two streams. In this sense, rule 3 of the previous section the streams.
describes a first-level loop breaking. 2) They are in a state in which removing one more unit would
In the case of four streams a second-level loop can exist. make compliance with the heating and cooling require-
Then the four streams are connected by four exchangers as ments impossible.
illustrated in Fig. 2. From the heat exchanger pairs aa/bb and These resulting networks are referred to as presolutions.
ab/ba respectively, the one with the smaller heat load can be They can have different numbers of units. They include those
eliminated. solutions that have the theoretical minimum number of units.
Not all presolutions are feasible because they can include heat
exchangers with negative freedoms.
Here, networks in which all exchangers have nonnegative
freedoms are called solutions. The freedom of an exchanger
depends on its position in the sequence of exchangers.
Figure 2. Second-level loop.
Therefore, the sequence of the exchangers in the presolutions
must be changed systematically to find out whether a
The heat loads of the four exchangers can be altered only
sequence without negative freedoms exists. Generally, there
simultaneously. If the heat loads of the exchangers aa and bb
is only a small number of solutions among the presolutions.
are altered by ± DQ, the heat loads of the exchangers ab and ba
There may be no solution at all.
change by DQ with the inverse sign. This follows simply from
During the course of development, the created intermedi-
the fact that the sum of the heat loads given or received in each
ate networks satisfy the first condition. The freedoms of the
of the streams must remain unchanged. If the value of DQ is
exchangers in the intermediate networks are not taken into
chosen equal to the negative (aa, bb) or positive (ab, ba) heat
account. Their changes of value, however, are calculated at
load of an exchanger, then this exchanger is eliminated. If the
every step because the freedoms must be known at the end.
exchangers of a pair have equal heat load, both are removed.
The method cannot change the given system. It is not
The alteration of the freedoms depends on the ratio r of the
possible to provide for stream splitting or to divide the system
heat capacity flow rates. Tab. 1 shows the changes which are
at a pinch-point. The values given by the task table are taken as
produced on the freedom Fh of the four exchangers in the form
they are.

Table 1. Change of freedoms of exchangers after second-level loop breaking


(Fig. 2).
4 Minimum Number of Units
exchanger aa ab ba bb
DFh 1 r 1
There is a rule which determines the minimum number of
r³1 0 +1
DQ r r
units necessary to fulfill the heating and cooling demands of
DFh 1 r
r³1 DQ r ±1 ‡ 1r 0 the streams in a network. Linnhoff et al. [12] give the formula:
Convention: the sign of DQ is negative when the heat load of the pair aa/bb is
reduced and positive when the pair ab/ba is concerned. Nu min = Nstr + Nut + Nloops ± Ncomp (11)

Chem. Eng. Technol. 25 (2002) 11, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1109 $ 17.50+.50/0 1109
Full Paper

where 5.1 Generating the System of Subnetworks

Nu min minimum number of units The TI method finds five subnetworks. Their sequence is
Nstr number of process streams compiled in Tab. 3.
Nut number of utility streams (hot or cold)
Nloops number of independent loops Table 3. Sequence of subnetworks at the cold end of TC3.

Ncomp number of separately balanced components in Temperatures [C] streams number of:
the network cold hot included exchs. var (nvar)

70.00 90.00
The entire system counts for a balanced component so that SN1 1, 2, 3, 4, 5 2 6
45.00 65.00
we have the relation SN2 1, 2, 4, 5 2 2
40.00 60.00
SN3 2, 4, 5 1 2
Ncomp ³ 1 (12) 31.67 51.67
SN4 4, 5 0 1
25.00 45.00
In the presolutions, the number of loops Nloops is always SN5 20.00 40.00 4 0 1
equal to zero and the number of utilities Nut can take only the
values one or zero because only a heating utility or a cooling
utility can exist.
The number of variants nvar in the subnetworks, as shown in
the last column of Tab. 3, is calculated by Eq. (1). The number
of main variants is Nvar main = 24 according to Eq. (2). In the
5 Explaining the Method with an Example present example, there is no main variant which includes all
connections between the streams.
As an example, the cold-end side of the problem TC3 [13] is Stream No. 3 occurs only in the first subnetwork. Therefore,
chosen. TC3 is a pinched problem, so both sides are treated the connections 3±4 and 3±5 can only be realized in this
separately. The cold-end side requires stream splitting. Tab. 2 subnetwork, but it is not possible to realize both because for
shows the task table of the problem. The streams number 2 and one stream only one heat exchanger is allowed. That means
3 are the split streams. The minimum temperature difference that all of the 24 main variants are missing at least one
is DTmin = 20 K. connection. Among these variants, there are two which are
missing exactly one connection (3±4 or 3±5). The others are
missing additional connections due to different coincidences
Table 2. Task table for the stream-split cold-end side of TC3 (case b in [13]). of combinations.
Because the proposed method requires a main variant
Ch/k Ts Tt Q which contains all connections between the streams, we must
Stream [kW/K] [C] [C] [kW] introduce an additional subnetwork in which the missing
1h 2.0 90.00 60.00 60
connection is realized in the form of an exchanger with zero
heat load.
2h 3.0 90.00 51.67 115
The further development is shown using the main variant in
3h 5.0 90.00 65.00 125 which only connection 3±4 is absent. Equally, one could have
4c 2.5 20.00 70.00 ±125 chosen a main variant with more than one missing connection.
5c 3.0 25.00 70.00 ±135
If a stream appears in more than one missing connection, then
more than one zero subnetwork must be generated because, as
mentioned, only one exchanger for every stream is allowed in
We see in the task table that the heat demand of stream No. 4 a subnetwork. The insertion of the zero subnetworks can be
is equal to the cooling demand of stream No. 3. That means done at the left-hand side or at the right-hand side of any such
that there is a balanced subset in the system. Then, according subnetwork in which both streams of the missing connection
to Eq. (12), we have Ncomp = 2. The total balance of the system appear.
is The insertion of a zero exchanger is not necessary if the
missing connection can be generated during the following step
300 kW ± 260 kW = + 40 kW of balancing heaters and coolers across the boundaries of
subnetworks. But it is not allowed to cross an inserted zero-
and, consequently, a cooling utility is needed which means Nut subnetwork when balancing heaters and coolers.
= 1. Then, according to Eq. (11), the minimum number of units In our example, we choose the insertion of the zero
becomes subnetwork at the left-hand side of SN1 because there, no
balancing of heaters and coolers across the boundary can
Nu min = 5 + 1 ± 2 = 4 occur. The resulting network is shown in Fig. 3.

1110 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1110 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 11
Full Paper

ED rules 1 and 2. The freedoms of a merged exchanger are


given by ED rule 3. Initially, all heat exchangers which had
been created within the subnetworks have the freedom zero
and the newly created ones have the freedom given by the
Eqs. (13).

5.3 Balancing Heaters and Coolers (First Recursive


Figure 3. Problem TC3, right-hand side of the pinch-point with stream splitting.
Algorithm)
Structure before dissolution of the subnetworks (one of 24 main variants).
In the next step, all heaters and coolers are balanced
through an exchanger following ED rule 4. Obviously, the
balancing is not possible if there is no heat exchanger between
5.2 Dissolution of the Subnetworks the corresponding hot and cold streams. At this point it
becomes evident that the premise of the existence of all heat-
As pointed out in the explanation of the TI method, the exchanging connections is necessary. Zero connections can
subnetworks include heaters and coolers which are necessary acquire a positive heat load here.
to complete the heat transfer tasks. The next step is to balance In our example, stream No. 1 needs no utility (see Fig. 3). So
the heating and cooling quantities across the boundaries of the the balancing takes place between the hot streams 2 and 3, and
subnetworks. The procedure of balancing creates a heat the cold streams 4 and 5. In the present stage of the
exchanger with a heat load Q equal to the smaller value of the development, heaters and coolers have been merged into
cooling and heating loads. Except when these heat loads are one unit in each stream, and we can balance them through the
equal, there remains an excess heating or cooling load. This is exchangers 2±4 and 3±5 or, alternatively, through the
to be placed in direction of the stream behind the created heat exchangers 2±5 and 3±4.
exchanger. The exchanger receives the freedoms Generally, the heat loads of the balanced heaters and
coolers are not equal and a rest remains. The remaining
if r > 1: Fc = Q (13a) heaters and coolers are again balanced in all possible variants.
This is repeatedly done in a recursive procedure until only
if r < 1: Fh = Q (13b) heaters or coolers remain, dependent on the surplus or
shortage of heat in the system. The result is a branching tree. A
If there is more than one heater or cooler at the SN branch terminates when there are no more heaters and coolers
boundary, different possibilities to balance sinks and sources to balance or when a generated system is equal to a previously
arise. Fig. 3 shows such a situation at the boundary of SN1 and generated system.
SN2, where two variants exist. The different possibilities lead In this way, a series of variants to the chosen subvariant are
again to a branching of variants, which we call subvariants to a created, which here are called sub-subvariants. As the given
main variant. If the main variant includes all connections or if procedure does not destroy connections, the sub-subvariants
missing connections are realized in zero subnetworks, every contain all connections and we can choose any of them for
subvariant includes all connections and any of them can be further evolution. Therefore, it is sufficient to develop only
chosen for further development. But in the case of connec- one of the branches.
tions which have been created by the balancing of heaters and In the example, four balancing matches are possible,
coolers at a SN boundary, there may be incomplete namely, 2±4, 2±5, 3±4 and 3±5. So the recursive procedure
subvariants. begins with four branches. It leads to five sub-subvariants. We
In the given example (Fig. 3), there exist only the variants at choose one of them arbitrarily to enter the loop-breaking
the boundary of SN1 and SN2, namely, 2±5 and 3±5. The match procedure. The structure of the chosen sub-subvariant is
2±4 at the boundary SN2/SN3 is without alternative. That shown in Fig. 4.
means that two subvariants exist. Because both contain all
connections, both can be chosen to proceed with the next
steps.
At this stage, the subnetworks are dissolved and their
boundaries have become meaningless. The next step is
merging identical units, i.e., heat exchangers between the
same streams and heaters or coolers in the same stream. To do
this, exchangers, coolers, and heaters must be shifted together,
which causes changes of the freedoms in accordance with Figure 4. State of the network before entering the loop-breaking procedure.

Chem. Eng. Technol. 25 (2002) 11, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1111 $ 17.50+.50/0 1111
Full Paper

5.4 Loop Breaking (Second Recursive Algorithm) certain heat-exchanging connections are desirable or others
are to be avoided. A presolution which is not a solution
The process of loop breaking has been explained in section contains one or more exchangers with negative freedoms. To
2.3. In the stage of evolution which is now reached, the get a solution, the heat exchanger with negative freedoms and
network is examined in order to find all second-level loops in the heat load Q must be decomposed into two exchangers with
the chosen sub-subvariant. Heaters or coolers (there can be the heat loads Qp and Qr :
only one or the other) are included in the search and are now
considered as heat exchangers with infinite freedoms. In order Q = Qp + Qr (14)
to reach the complete range of results, all hot or cold streams
respectively must have a heat-exchanging connection to the The exchanger with the heat load Qp is substituted by a
utility stream. If this is missing in a process stream, it is inserted heating and a cooling in the participating streams and thus
with the heat load zero. This procedure corresponds to the deleted (Fig. 5). Qr is the heat load of the remaining
inserting of zero heat exchangers. But here it is not necessary exchanger. Qp is equal to the energy penalty and corresponds
to consider the freedoms so that the insertion can be done to the smaller value of the two freedoms, i.e.,
anywhere.
Here, a loop is always a second-level loop. First-level loops Qp = ± Fh if r > 1 (15a)
do not occur; they have been removed before. Higher-level
loops will be broken automatically when the second-level Qp = ± Fc if r < 1 (15b)
loops are broken in sequence. Every loop breaking has two
solutions. So, n loops in an initial system produce 2.n new
networks. These networks can again contain loops. Corre-
sponding to the procedure of balancing heaters and coolers, a
recursive procedure is carried out which terminates when no
more loops are found or when an identity occurs. A system
without loops is a presolution.
The network of our example is completed with the zero
Figure 5. Decomposition of a heat exchanger with negativ freedoms (case r > 1).
coolers 1±0 and 3±0. The stream of the cooling utility is
included as stream number 0. We find nine second-level loops
in the network, beginning with the sequence 1±4, 1±5, 2±4, 2±5. If the heater and the cooler are placed at the right-hand
side of the exchanger in the case r > 1 and at the left-hand
side when r < 1, the freedoms of the exchanger remain
5.5 Solutions unchanged after the decomposition. After shifting the heater
to the hot end and the cooler to the cold end of the network,
Finally, the sequence of the heat exchangers in the the freedoms of the remaining exchanger become zero. The
presolutions is systematically changed in all combinations. resulting system contains at least one unit more than the
In this process, the changes of the freedoms follow the ED minimum number.
rules 1 and 2. If the freedoms of all exchangers in the sequence In complex systems it can happen that the added utilities
are zero or positive, then the presolution is a solution. form loops with heat exchangers in the network. So, one can
Heaters or coolers do not participate in the change of reduce the number of units by loop breaking. But obviously, it
sequence. Coolers are placed at the right-hand side and is not allowed to break the loop when the decomposed
heaters at the left-hand side of the system. In this way the exchanger is involved.
freedoms have the highest possible level. In the system resulting from loop breaking one often finds
The chosen example leads to two presolutions with the that the exchanger with the remaining heat load has changed
minimum number of four units. Both of them are solutions, its freedoms from zero to positive values. That means that the
and one of them is a solution given by Linnhoff and introduced energy penalty is higher than necessary. In this
Hindmarsh [13]. The authors of [13] give a second solution case, a new calculation can be carried out with a reduced
to the problem TC3. This cannot be obtained here because its energy penalty. It is not uncommon that even after reducing
target temperatures of the split streams are different. There- the energy penalty a heat exchanger with originally negative
fore, the solutions belong to different problems in view of the freedoms still shows freedoms which are greater than zero so
method proposed here. that a new calculation is necessary. One can avoid such
repeated calculations by using a method proposed by Mehta et
al. [14]. The authors use a graph representation which allows
6 Upgrading Presolutions to determine the optimal heat load distribution between the
remaining heat-exchanging connections after deletion of one
Every presolution can be made a solution if an energy exchanger. So, the network with the minimum energy penalty
penalty is accepted. The reason for this can be for instance that can be found.

1112 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1112 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 11
Full Paper

7 Survey of Solutions for Some Problems Reported number of presolutions and solutions. Program runs in which
in the Literature the search is restricted to networks with predefined connec-
tions lead to the six solutions given by [1,12,19±21] immedi-
A series of problems taken from the literature has been ately.
calculated with a system of C++ computer programs. Tab. 4
shows the results.
The examples of [13,15±17,19] are theoretically con- 8 Conclusion
structed. The example taken from [18] represents the heat
exchanger network of an oil refinery. The number of main Although the method proposed in this paper does not offer
variants is given for information only. Based on the conjecture the possibility of optimizing a HEN structure, it can be a useful
introduced in section 2.1, only one of the variants is used. tool for finding the complete range of solutions for a given
Decisive for the calculation time is the number of steps in the problem as part of a more extensive procedure. Additionally,
recursive loop-breaking procedure. This number is the the method provides a good insight into the construction of the
difference between the number of units remaining after networks.
balancing heaters and coolers (section 5.3) and the minimum
number of units. The number of recursive steps is 11 for the
problem of [16] with 8 streams and 16 for the one of [18] with Acknowledgement
10 streams. For these problems it was not possible to follow all
paths of the branching tree within a reasonable calculation The author wishes to thank his son, Dr. R. Schack, for advice
time. Therefore, the search was restricted to two second-level in C++ programming.
loops in each step. When the calculation was started with Received: September 19, 2001 [CET 1492]
different loops in the first step, the range of resulting
presolutions remained the same. So, it is not unlikely that
the results include all presolutions. Symbols used
In the case of the problem 10SP1, the number of recursive
steps is 17 whithout introducing zero coolers, which com- Cc, Ch [kW/K] heat capacity flow rate
pletely rules out a systematic search. But there is a great Fc, Fh, Fs [kW] freedom

Table 4. Examples calculated.

DTmin [K] Nstr Nu min Nsn Nvar main Number of Number of


presolutions solutions

4SP1 [15] r 11.11 4 4 6 8 4 2

SundØn [16] r,z 10.00 4 4 3 4 6 1

5SP1 [17] u 11.11 5 5 8 144 18 5


1)
TC3 [13] r,s,z 20.00 5 4 5 24 2 2

6SP1 [15] u 11.11 6 6 10 3456 78 2

Fa.&Su. [18] r,z 30.00 6 6 4 96 45 3

7SP1 [17] u 11.11 7 7 11 20736 240 2

7SP2 [17] u,z 11.11 7 7 12 5184 243 56


2)
SundØn [16] l,s 10.00 8 7 7 955328 54 92)

Fa.&Su. [18] l,s,z 30.00 10 7 10 1492992 22) 22)

10SP1 [19] u 11.11 10 10 17 4.953 ´ 1015 3) 3)

r: right hand of pinch-point (cold end)


l: left hand of pinch-point (hot end)
u: unpinched problem
s: system with stream splitting (additional streams resulting from splitting are included in the number of streams Nstrr)
z: zero subnetwork introduced (not included in the number of subnetworks Nsn)
1)
stream splitting case b in [13]
2)
possibly incomplete
3)
not determined

Chem. Eng. Technol. 25 (2002) 11, Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1113 $ 17.50+.50/0 1113
Full Paper

Nsubscr [±] number of (subscript) Subscripts (other)


Q [kW] heat load
Qc, Qh [kW] quantity of external cooling of heating c cold stream
Qp [kW] energy penalty h hot stream
Qr [kW] remaining heat load after s smaller of two values (c,h)
decomposition of an exchanger
Ts [C] supply temperature
Tt [C] target temperature References
nsl [±] the larger number of hot or cold
streams [1] B. Linnhoff, J. R. Flower, AIChE J. 1978, 24 (4), 642.
[2] N. Nishida, G. Stephanopoulos, A. W. Westerberg, AIChE J. 1981, 27
nss [±] the smaller number of hot or cold (3), 321.
streams [3] J.-L. Su, R. L. Motard, Comput. Chem. Eng. 1984, 8 (2), 67.
nvar [±] number of variants within a subnetwork [4] J. Calandranis, G. Stephanopoulos, Chem. Eng. Res. Des. 1986, 64, 347.
[5] C. A. Floudas, A. R. Ciric, I. E. Grossmann, AIChE J. 1986, 32 (2), 276.
r [±] ratio of hot and cold heat capacity flow [6] K. K. Trivedi, B. K. O'Neill, J. R. Roach, R. M. Wood, Comput. Chem.
rates within an exchanger Eng. 1990, 14 (6), 601.
[7] S. Chakraborty, P. Ghosh, Chem. Eng. J. 1999, 72 (3), 209.
DTmin [K] minimum allowable temperature [8] Y. Ren, B. K. O'Neill, J. R. Roach, Ind. Eng. Chem. Res. 2001, 40 (4),
difference within an exchanger 1168.
DTs [K] smallest effective temperature [9] B. Linnhof, J. R. Flower, AIChE J. 1978, 24 (4), 633.
[10] R. Schack, personal communication.
difference within an exchanger [11] K. Schack, Chem. Ing. Tech. 1992, 64 (11), 1031.
[12] B. Linnhoff, D. R. Mason, I. Wardle, Comput. Chem. Eng. 1979, 3 (3),
295.
[13] B. Linnhoff, E. Hindmarsh, Chem. Eng. Sci. 1983, 38 (5), 745.
Subscripts (to N) [14] R. K. C. Mehta, S. K. Devalkar, S. Narasimhan, Trans. Ichem. E. 79 Part
A 2001, 143.
[15] K. F. Lee, A. H. Masso, D. F. Rudd, Ind. Eng. Chem. Fundam. 1970,
comp separately balanced components 9 (1), 48.
sn subnetworks [16] B. SundØn, Optimal Design of Heat Exchanger Networks, Chalmers
str process streams Tekniska Högskola, Publ. No. 87/1, Göteborg 1987.
[17] A. H. Masso, D. F. Rudd, AIChE J. 1969, 15 (1), 10.
u min minimum of units [18] B. Farhanieh, B. SundØn, Heat Recovery Systems & CHP 1990, 10 (3),
ut utilities 285.
var main main variants [19] T. K. Pho, L. Lapidus, AIChE J. 1973, 19 (6), 1182.
[20] N. Nishida, Y. A. Liu, L. Lapidus, AIChE J. 1977, 23 (1), 77.
[21] S. A. Papoulias, I. E. Grossmann, Comput. Chem. Eng. 1983, 7 (6), 707.

_______________________

1114 Ó 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 0930-7516/02/1111-1114 $ 17.50+.50/0 Chem. Eng. Technol. 25 (2002) 11

You might also like