Title: An efficient global-local hybrid optimisation method using design sensitivity analysis

Authors: Kyu-Yeul Lee, Seonho Cho, Myung-II Roh

Addresses: Research Institute of Marine Systems Engineering and Department of Naval Architeture and Ocean Engineering, Seoul National University, San 56-1, Shilim-Dong, Kwanak-ku, Seoul, 151-742, Korea. Research Institute of Marine Systems Engineering, Seoul National University, San 56-1, Shilim-Dong, Kwanak-Ku, Seoul, 151-742, Korea. Departiment of Naval Architecture and Ocean Engineering, Seoul National, University, San 56-1, Shilim-Dong, Kwanak-Ku, Seoul, 151-742, Korea

Abstract: An efficient global-local hybrid optimisation method is developed combining a genetic algorithm (GA) as a global optimiser and a method of feasible direction (MFD) as a local optimiser. During the global iteration using the GA, the design sensitivity information with respect to design variables is used to perform a what-if study for efficient crossover. After the final global iteration, a local optimisation can be followed to further improve the solution obtained from the GA. The developed global-local hybrid optimisation method is applied to simple mathematical problems to compare the efficiency and accuracy of the proposed method with that of the conventional genetic algorithm. The application of the developed method is also extended to the practical engineering problems, which shows very good and efficient results.

Keywords: global-local hybrid optimisation method, design sensitivity analysis.

DOI: 10.1504/IJVD.2002.001992

International Journal of Vehicle Design, 2002 Vol.28 No.4, pp.300-317

Published online: 15 Aug 2003 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article