Page 284 - Design and Operation of Heat Exchangers and their Networks

270 Design and operation of heat exchangers and their networks

problems. In the master-level optimization, an evolutionary method was
developed for generating network configurations that are then sent to the
slave level for continuous variable optimization using a memetic particle
swarm optimization algorithm. Their case studies showed an excellent
search ability of the bilevel algorithm.

6.5.4 Knowledge-based expert system

Recently, there has been a very rapid growth in the application of
knowledge-based system in synthesis of heat exchanger networks. For a
given network design problem, there are large numbers of possible config-
urations, and the optimal solution cannot be obtained by a simple search
method. An efficient method is to establish a mathematical model of net-
work superstructure. For the available superstructures, there is possibility
for each hot stream and each cold stream to match; the superstructure will
be huge in scale if number of hot and cold streams exceeds a certain number,
which makes the search for optimal solution complicated and difficult to be
obtained, especially when the number of streams is large. Furthermore, in
the whole process of simultaneous synthesis of heat exchanger networks
by adopting stagewise superstructure, the network configuration was kept
as a black box, and experienced experts were kept outdoor in the process
and cannot help with the scheme choice. In addition, the giant search scope
makes the calculation a complete and arduous job. While from the view-
point of engineering application, some of hot stream and cold stream
matches involved in the superstructure are not necessary; on the other hand,
not every possible structure is contained in those superstructures. An expert
system can help to present a more reasonable and relatively simple super-
structure to make search for the optimal solution easy to realize.
The expert system method is an artificial intelligence technique. The
principles in expert system method are a set of rules based on logical knowl-
edge and experiences such as whether the stream should be split or not, how
to determine of split number if the split of a stream is necessary, and the
match principle of streams. These principles are translated into computer
language in the calculation model.
Generally, an expert system consists of a database, a knowledge base,
control strategies, and a man-machine interface. The database is divided into
two parts: static and dynamic bases. The static base records the initially
assigned data and relevant data required for solving the problem, and the
dynamic base stores all the intermediate information generated while solving