Evolutionary Design and Multi–objective Optimisation

In this paper we explore established methods for optimising multi-objective functions whilst addressing the problem of preliminary design. Methods from the literature are investigated and new ones introduced. All methods are evaluated within a collaborative project for whole system airframe design and the basic problems and difficulties of preliminary design methodology are discussed (Cvetković, Parmee and Webb 1998). Our Genetic Algorithm is expanded to integrate different methods for optimising multi–objective functions. All presented methods are also analysed in the context of whole system design, discussing their advantages and disadvantages. The problem of qualitative versus quantitative characterisation of relative importance of objectives (such as ‘objective A is much more important then objective B’) in multi–objective optimisation framework is also addressed and some relationships with fuzzy preferences (Fodor and Roubens 1994) and preference ordering established. Several new predicates between objectives ( , , etc.) are introduced (Cvetković and Parmee 1998). Some examples of application in GA based multi–objective framework are also presented.