Title: Multiobjective design of a vehicular structure using metamodelling and an efficient genetic algorithm

Authors: Hongbing Fang, Qian Wang

Addresses: Department of Mechanical Engineering and Engineering Science, University of North Carolina at Charlotte, 9201 University City Boulevard, Charlotte, NC 28223-0001, USA. ' Skidmore, Owings and Merrill LLP, One Front Street, San Francisco, CA 94111, USA

Abstract: A new Non-dominated Sorting (NS) algorithm is presented in this paper to improve the efficiency of Genetic Algorithms (GAs) in solving multiobjective optimisation problems. Commonly used NS algorithms typically have a time complexity of O(MN²) to obtain a set of N non-dominated solutions for M objectives. In the proposed NS algorithm, a new dominance tree structure and a divide-and-conquer mechanism are adopted to reduce the number of redundant comparisons in determining non-dominated solutions. The new algorithm is implemented into Non-dominated Sorting Genetic Algorithm (NSGA)-II and shown to have improved overall efficiency of the entire evolution processes. By combining with the metamodelling technique, the new algorithm is successfully used in the multiobjective design of a vehicular structure for safety improvement and weight reduction.

Keywords: multiobjective design; optimisation; metamodelling; radial basis functions; RBFs; Pareto frontier; genetic algorithms; GA; non-dominated sorting; design engineering; vehicle structure; vehicle safety; weight reduction; vehicle design.

DOI: 10.1504/IJDE.2007.015334

International Journal of Design Engineering, 2007 Vol.1 No.1, pp.41 - 55

Published online: 10 Oct 2007 *

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