12-03-2014, 04:52 PM
Performance of the improved artificial bee colony algorithm on standard engineering constrained problems
Performance of the improved.pdf (Size: 688.09 KB / Downloads: 34)
[b]Abstract[/b]
Artificial bee colony (ABC) algorithm is successfully
used for many hard, mostly continuous, optimization problems. There
is a way to extend standard ABC algorithm to constrained problems.
In this paper an improved version of the artificial bee colony
algorithm adjusted for constrained optimization problems is
presented. It uses Deb’s rule. This modified algorithm has been
implemented and tested on four standard engineering constrained
benchmark problems which contain discrete and continuous
variables. Our results were compared to the results obtained by
simple constrained particle swarm optimization algorithm (SiC-PSO)
which showed a very good performance when it was applied to the
same problems. Our results are of the comparable quality with faster
convergence.
INTRODUCTION
CONSTRAINED optimization problems have numerous
applications. Engineering design is one of the scientific
fields in which constrained optimization problems frequently
arise [1]. These types of problems normally have mixed
(continuous and discrete) design variables, nonlinear objective
functions and nonlinear constrains. Constrains are very
important in engineering design problems. They are usually
imposed in the statement of the problems and sometimes are
very hard to satisfy, which makes the search difficult and
inefficient.
PROPOSED ALGORITHM: SC-ABC
In our proposed approach (called Simple Constrained
Artificial Bee Colony, or SC-ABC) as in the ABC for
constrained problems, algorithm uses Deb’s rules instead of
the greedy selection in order to decide what solution will be
kept. The expressions for evaluating probability Eq. (1), for
producing a candidate food position Eq. (2) and for
initialization new food sources Eq. (3) stayed the same as in
the version of the ABC proposed for unconstrained
optimization problems.
CONCLUSION
In this paper, we present an improved ABC algorithm for
constrained problems (SC-ABC). The SC-ABC was tested on
three constrained optimization problems which contain
discrete and continuous variables. The algorithm showed a
good performance. We compared our results to the results
reached by Simple Constrained Particle Swarm optimization
algorithm (SiC-PSO) which showed a very good performance
when it was applied to the same problems. Although our
algorithm did not obtain the optimal values for each tested
problem, the average values reached by SC-ABC are better.
We can conclude that the SC-ABC can quickly search toward
the global optimum and can be a promising alternative for
solving this sort of problems due to its simplicity and
reliability. As part of our future work, we are interested to
perform a more detailed statistical analysis of the performance
of our proposed approach and to improve the new algorithm's
ability to escape the local attractors.