Authors: Jongsoo Lee, Seungjin Kim, Shinill Kang
Addresses: Department of Mechanical Engineering, Yonsei University, Seoul, 120-749, Korea. Department of Mechanical Engineering, Yonsei University, Seoul, 120-749, Korea. Department of Mechanical Engineering, Yonsei University, Seoul, 120-749, Korea
Abstract: This paper describes the construction of global function approximation models for use in design optimisation via global search techniques such as genetic algorithms. Evolutionary fuzzy modelling (EFM) is implemented in the context of global approximate optimisation. Such approximation methods may have their promising characteristics in a case where the training data is not sufficiently provided or uncertain information may be included in the design process. Fuzzy inference system is central to identifying the input-output relationship in both methods. This paper introduces the general procedures including fuzzy rule generation, membership function selection and inference process in EFM, and presents its generalisation capabilities in terms of the number of fuzzy rules and training data. A three-bar truss design is first considered as a benchmark, and sizing of automotive A-pillar with rib structures for passenger protection is further explored in this context of EFM-based approximate optimisation.
Keywords: global approximate optimisation; responses surface method; fuzzy modelling; genetic algorithms; A-pillar trim design.
International Journal of Vehicle Design, 2002 Vol.28 No.4, pp.339-355
Available online: 15 Aug 2003 *Full-text access for editors Access for subscribers Purchase this article Comment on this article