Abstract: (8206 Views)
An efficient selection operator for use in genetic search of pipe networks optimal design
is introduced in this paper. The proposed selection scheme is the superior member of a family of
improved selection operators developed in an attempt to more closely simulate the main features of
the natural mating process which is not reflected in existing selection schemes. The mating process
occurring in the nature exhibits two distinct features. First, there is a competition between
phenotypes looking for the fittest possible mate which usually ends up with choosing a mate with
more or less the same fitness. Second, and more importantly, the search for a mate is often confined
to a community of phenotypes rather than the whole population. Four different selection operators
simulating these features in a random and pre-determined manner are developed and tested. All the
selection schemes exhibit good convergence characteristics, in particular the one in which both the
size of the sub-community and the pair of the mates in the sub-community are determined randomly.
The efficiency of the proposed selection operator is shown by applying the method for the optimal
design of three well-known benchmark networks, namely two-loop, Hanoi and New-York networks.
The proposed scheme produces results comparable to the best results presented in the literature
with much less computational effort
Type of Study:
Research Paper |