Computational Technology Resources - CCP

Keywords: optimisation, evolution, genetic algorithm, particle swarm optimisation, efficiency, performance comparison, population size, truss.

Summary

Genetic algorithms (GAs) and particle swarm optimisation (PSO) are well-known for their ability in obtaining global optima. Some evidence exists in the structural engineering literature that PSO involves less overall computation effort than a GA. Hence, these two methods have been selected and benchmarked against each other to test their relative robustness and efficiency for structural optimisation applications. This paper examines the performance and efficiency of these two optimisation algorithms in solving both mathematical benchmark functions and the classical ten-bar truss redundant problem. The mathematical functions used are a ten-dimensional Rosenbrock valley function and a ten-dimensional Schwefel function. These have many local minima and each has a unique and known global minimum. Tests are performed to assess the performance of each in relation to the population size required and the number of generations to achieve convergence. The GA produces slightly better performance for Rosenbrock's function while the PSO works best for Schwefel's function. Both methods find the correct solutions when enough iterations are allowed. The PSO performs best when a limited population or limited iterations are used. When the ten-bar redundant truss problem is considered, the PSO clearly outperforms the GA for all categories tested. To examine the efficiency of these two algorithms, population sizes of 10, 50, 100, 150 and 200 were used. The best results at iteration counts 100, 200, 400 and 800 were logged. Both methods are sensitive to the population size. The larger the number of candidates in the population, the better the solution for both the GA and PSO. PSO seems to perform better than the GA for smaller populations. Therefore, for the more complex problems, the PSO is shown to outperform the GA for smaller population sizes and limited iterations.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £65 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Proceedings ISSN 1759-3433 CCP: 99 PROCEEDINGS OF THE ELEVENTH INTERNATIONAL CONFERENCE ON COMPUTATIONAL STRUCTURES TECHNOLOGY Edited by: B.H.V. Topping Paper 68 A Comparison of Genetic Algorithm and Particle Swarm Optimisation for Theoretical and Structural Applications Z. Wang, T.J. McCarthy and M.N. Sheikh Faculty of Engineering, University of Wollongong, Australia doi:10.4203/ccp.99.68 purchase the full-text of this paper Full Bibliographic Reference for this paper Z. Wang, T.J. McCarthy, M.N. Sheikh, "A Comparison of Genetic Algorithm and Particle Swarm Optimisation for Theoretical and Structural Applications", in B.H.V. Topping, (Editor), "Proceedings of the Eleventh International Conference on Computational Structures Technology", Civil-Comp Press, Stirlingshire, UK, Paper 68, 2012. doi:10.4203/ccp.99.68 Keywords: optimisation, evolution, genetic algorithm, particle swarm optimisation, efficiency, performance comparison, population size, truss. Summary Genetic algorithms (GAs) and particle swarm optimisation (PSO) are well-known for their ability in obtaining global optima. Some evidence exists in the structural engineering literature that PSO involves less overall computation effort than a GA. Hence, these two methods have been selected and benchmarked against each other to test their relative robustness and efficiency for structural optimisation applications. This paper examines the performance and efficiency of these two optimisation algorithms in solving both mathematical benchmark functions and the classical ten-bar truss redundant problem. The mathematical functions used are a ten-dimensional Rosenbrock valley function and a ten-dimensional Schwefel function. These have many local minima and each has a unique and known global minimum. Tests are performed to assess the performance of each in relation to the population size required and the number of generations to achieve convergence. The GA produces slightly better performance for Rosenbrock's function while the PSO works best for Schwefel's function. Both methods find the correct solutions when enough iterations are allowed. The PSO performs best when a limited population or limited iterations are used. When the ten-bar redundant truss problem is considered, the PSO clearly outperforms the GA for all categories tested. To examine the efficiency of these two algorithms, population sizes of 10, 50, 100, 150 and 200 were used. The best results at iteration counts 100, 200, 400 and 800 were logged. Both methods are sensitive to the population size. The larger the number of candidates in the population, the better the solution for both the GA and PSO. PSO seems to perform better than the GA for smaller populations. Therefore, for the more complex problems, the PSO is shown to outperform the GA for smaller population sizes and limited iterations. purchase the full-text of this paper (price £20) go to the previous paper go to the next paper return to the table of contents return to the book description purchase this book (price £65 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions