Professional Documents
Culture Documents
www.elsevier.com/locate/compchemeng
Abstract
A conceptual tool is presented to address the flexibility and operability objectives for heat exchanger networks. The approach
presents a screening model to accommodate the considerations ahead of design. The new development is essentially a targeting
tool based on the Area Target Model that is a part of the Hypertarget methodology of (Chem. Engng. Sci., 54 (1999) 519; Chem.
Engng. Sci., 54 (1999) 541; Chem. Engng. Sci., 54 (1999) 685). The targets accommodate for an early pre-processing of matches
and assess implications due to parameter variations. The feasibility targets provide ahead of design information on the costs
incurred due to the variations. Flexibility targets assess the ability of the system to handle variations for different driving forces.
The potential of the matches to handle variations can be further exploited with superstructure schemes. The paper proposes a
systematic algorithmic procedure that converges to flexible networks at the expense of only a minor set of additional constraints.
Examples are presented to illustrate the approach, explain its ability to assess the impact of variations and prove the ease of its
use in real-size problems. © 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Feasibility targets; Fouling; Pinch analysis; Flexibility targets; Heat transfer coefficients
0098-1354/01/$ - see front matter © 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 9 8 - 1 3 5 4 ( 0 1 ) 0 0 6 5 3 - 6
888 L. Tantimuratha et al. / Computers and Chemical Engineering 25 (2001) 887–892
several types and purposes. Most notably, one should 2.2. Outline of de6elopments
refer to the MILP transhipment model by Papoulias
and Grossmann (1983) and the NLP superstructure by The new developments include:
Floudas, Ciric and Grossmann (1986). The former con- 1. conceptual models to screen layouts for grassroots
tributed with a conceptual model able to identify design;
matches of process streams; the latter extended these 2. conceptual models to screen modifications for
developments with superstructure schemes one could retrofit applications;
employ to automatically develop energy integrated net- 3. design procedures for the development of layouts of
works. However, the merits of Pinch Technology had networks with the ability to handle variations.
enabled the method to continue its ascendancy over Models in (1) and (2) apply prior to the development
mathematical programming in the vast majority of of a HEN network. They are formulated as MILP
industrial applications. Its ability to handle large size models and they measure the ability of the primal set to
problems and its potential to transparently explain handle variations at different levels of energy recovery
economic trade-offs has maintained a convincing edge (DTmin). The result of the optimisation curves can be
amongst practitioners and engineers. In due recognition plotted as flexibility targets alongside the Hypertargets.
of the power of thermodynamics, a new generation of Good designs are located in areas of large flexibility
applications illustrated venues to regain all the merits of targets and low total cost (i.e. Hypertargets).
the Pinch Method within a mathematical programming The network development stage represents an addi-
framework. tional challenge. The selected designs from (1) to (2) do
Hypertargets (Briones & Kokossis, 1999a,b,c) com- not represent networks yet. They are simply combina-
bine the merits of thermodynamics and mathematical tions of matches with a promising potential to account
programming and introduce a systematic framework for assumed variations in the design parameters. For
for HEN synthesis and retrofit applications. Hyper- guiding the optimisation search to exploit the potential
targets can be used to provide targets ahead of design of these matches, the mathematical models need be
in the form of a solution stream of design options expanded with a significant amount of additional mod-
available at different driving forces. A library of target- elling information. Instead the paper outlines an itera-
ing models is introduced (ATM, TAME, HEAT) one is tive procedure that requires only incremental
able to launch with minimal effort either to enable the modifications in the model and converges in practice
development of targets or to set up superstructures for within very few iterations.
layout optimisation. The screening models combine in-
formation about the number of matches, the level of
energy recovery and the potential savings on capital 3. Screening model for targeting
investment (heat transfer area, re-piping,
modifications). The new developments are based on the ATM and
TAME models developed by Briones and Kokossis
(1999a,b,c). The models consist of targeting terms for
2. Motivation and the outline of the approach the heat transfer area. They contain energy balances for
different temperature intervals and involve residuals
2.1. Moti6ation that reduce to the minimum heat duties required. The
participation of residuals in re-formulations with an
The application of the ATM (Briones & Kokossis additional consideration of flexibility (Swaney & Gross-
1999a) designates HEN layouts that are attractive from mann, 1985a,b) results in MILP problems with a large
an economic viewpoint. The experience with the use of number of variables. In the new formulation:
the model indicates that there exist numerous combina- 1. all constraints that contain residuals are relaxed;
tions of matches that feature similar — if not identical 2. an additional set of inequality constraints accounts
— performance. The select pool of the design candi- for physical bounds on the available energy to
dates comprises the primal set of the ATM. The de- recover.
tailed optimisation of the candidates eventually Heat balances for hot stream and hot utilities:
determines configurations that, although of similar cost,
perform quite differently in their ability to adjust to iH, % Qij = Q H
i . (1)
j MA
process variations and parameter changes. The idea to
expand the ATM model has been motivated by the Heat balances for cold stream and cold utilities:
large size of the primal sets in the majority of industrial
applications. Rather than processing all candidates, it jC, % Qij = Q C
j . (2)
MA
would be more efficient to consider only the ones with
an additional potential to handle variations. Heat contents of each hot stream i:
L. Tantimuratha et al. / Computers and Chemical Engineering 25 (2001) 887–892 889
Table 1
Stream data for example 1
Table 2
Feasibility and Flexibility targets of each options for example 1
Fig. 3. The algorithm for network optimisation with the similarity constraints.
3. Fig. 2c explains the flexibility targets against DTmin. inequalities that activate identical connections in the
Fig. 2d plots the cardinality of the primal set. It is superstructures. The iterative procedure is schematically
interesting to see that for small DTmin, the shown in Fig. 3.
deviation from the minimum-number-of-units rule
(Pinch Analysis). 4.1. Example 3
The cost of the network is 6690 k$ per year (ie. within References
the range of Hypertargets) but it fails to handle the
expected variations in the parameters. The results from Briones, V., & Kokossis, A. (1999a). Hypertargets: a conceptual
the iterative scheme that is proposed iterative earlier programming approach for the optimisation of industrial heat
exchanger networks-I. Grassroots design and network complexity.
yields the network of Fig. 4b. The procedure converges Chemical Engineering Science, 54, 519– 539.
in two iterations. The network (of the same combina- Briones, V., & Kokossis, A. (1999b). Hypertargets: a conceptual
tion) features a different structure from the network A programming approach for the optimisation of industrial heat
with a very small penalty in the total cost (1.9%). exchanger networks-II. Retrofit design. Chemical Engineering Sci-
Moreover, the network can afford all of the variations. ence, 54, 541– 561.
Briones, V., & Kokossis, A. (1999c). Hypertargets: a conceptual
programming approach for the optimisation of industrial heat
exchanger networks-III. Industrial applications. Chemical Engi-
5. Conclusion neering Science, 54, 685– 706.
Floudas, V. A., Ciric, A. R., & Grossmann, I. E. (1986). Automatic
The paper outlines a conceptual tool that is able to synthesis of optimum heat exchanger network configurations.
address flexibility objectives for heat exchanger net- American Institute of Chemical Engineering Journal, 32 (2), 276–
290.
works. The work extends the screening models (ATM/
Grossmann, I. E., & Floudas, C. A. (1987). Active constraint strategy
TAME) of the Hypertarget methodology (Briones & for flexibility analysis in chemical processes. Computers and
Kokossis, 1999a,b,c) to enable variations in the process Chemical Engineering, 11 (6), 675– 693.
parameters. The new models retain the screening nature Halemane, K. P., & Grossmann, I. E. (1983). Optimal process design
of the early developments in their ability to identify under uncertainty. American Institute of Chemical Engineering
Journal, 29 (3), 425– 433.
options ahead of detailed calculations. Flexibility
Papoulias, S. A., & Grossmann, I. E. (1983). A structural optimiza-
targets are calculated alongside the Hypertargets and tion approach in process analysis-II: heat recovery networks.
are useful to assess trade-offs between the design flexi- Computers and Chemical Engineering, 7 (6), 707– 721.
bility and economics. Moreover, the optimisation ap- Swaney, R. E., & Grossmann, I. E. (1985a). An index for operational
proach is enabled with an iterative scheme to materialise flexibility in chemical process design: part I-formulation and
the targets of the screening stage. The procedure opti- theory. American Institute of Chemical Engineering Journal, 31 (4),
621– 630.
mises the network layout using superstructure develop-
Swaney, R. E., & Grossmann, I. E. (1985b). An index for operational
ments. Although it addresses a much larger problem, flexibility in chemical process design: part II-computational al-
the additional synthesis effort is very limited: in most gorithms. American Institute of Chemical Engineering Journal,
cases, the scheme converges within two iterations. 31 (4), 631– 641.