Facial reconstruction

Search LJMU Research Online

Browse Repository | Browse E-Theses

Predetermined time constant approximation method for optimising search space boundary by standard genetic algorithm

Rajarathianm, K and Gomm, JB and Jones, KO and Yu, DL (2015) Predetermined time constant approximation method for optimising search space boundary by standard genetic algorithm. In: ACM International Conference Proceeding Series , 1008. pp. 38-45. (16th International Conference on Computer Systems and Technologies, 25th - 26th June 2015, Dublin Ireland).

[img] Text
CompSysTech_2015 Research Paper (86 Paper).pdf - Published Version

Download (875kB)

Abstract

In this paper, a new predetermined time constant approximation (Tsp) method for optimising the search space boundaries to improve SGAs convergence is proposed. This method is demonstrated on parameter identification of higher order models. Using the dynamic response period and desired settling time of the transfer function coefficients offered a better suggestion for initial Tsp values. Furthermore, an extension on boundaries derived from the initial Tsp values and the consecutive execution, brought the elite groups within feasible boundary regions for better exploration. This enhanced the process of locating of the optimal values of coefficients for the transfer function. The Tsp method is investigated on two processes; excess oxygen and a third order continuous model with and without random disturbance. The simulation results assured the Tsp method's effectiveness and flexibility in assisting SGAs to locate optimal transfer function coefficients. Copyright © 2015 ACM.

Item Type: Conference or Workshop Item (Paper)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
T Technology > T Technology (General)
Divisions: Electronics and Electrical Engineering
Publisher: ACM
Date Deposited: 14 Sep 2016 11:11
Last Modified: 14 Sep 2016 11:11
DOI or Identification number: 10.1145/2812428.2812466
URI: http://researchonline.ljmu.ac.uk/id/eprint/4077

Actions (login required)

View Item View Item