ANZIAM J. 46(E) ppC89--C101, 2005.

Comparing genetic algorithms and particle swarm optimisation for an inverse problem exercise

C. R. Mouser

S. A. Dunn

(Received 26 October 2004, revised 1 February 2005)

Abstract

We describe the performance of two population based search algorithms (genetic algorithms and particle swarm optimisation) when applied to the optimisation of a structural dynamics model. A significant difficulty arises when trying to compare the performance of such algorithms. For the two algorithms to perform at their best, several properties (for example, population size and mutation rate) need to be set. The performance of the algorithms can be highly sensitive to the choice of these parameters, and the optimisation of these leads to a search in a multi-dimensional space. This work describes how a genetic algorithm optimises the properties of a genetic algorithm and a particle swarm optimisation in order to produce algorithms that are optimally tuned to the particular problem being solved. The two methods are rigorously compared. This problem is implemented on a distributed computing facility spread across the Defence Science & Technology Organisation's network across four cities in south-east Australia.

Download to your computer

Authors

C. R. Mouser
S. A. Dunn
Defence Science & Technology Organisation, Melbourne, Australia. mailto:carl.mouser@dsto.defence.gov.au

Published March 21, 2005. ISSN 1446-8735

References

  1. S. A. Dunn, Optimised Structural Dynamic Aircraft Finite Element Models Involving Mass, Stiffness and Damping Elements. {International Forum on Aeroelasticity and Structural Dynamics}, Madrid, Spain, June 2001, pp.387--396
  2. J. Kennedy and R. C. Eberhaurt, Particle Swarm Optimization. {Proc. IEEE Int. Conf. Neural Networks}, Perth, Australia Nov. 1995, pp.1942--1948
  3. S. Dunn, S. Peucker and J. Perry, Genetic Algorithm Optimisation of Mathematical Models Using Distributed Computing. {Applied Intelligence}, in-press