The type of design problems considered have following characteristics:
Models for design structure, functionality and quantitative behavioural analysis are available.
Generative design relations are usually not available, while available adaptive relations are imprecise and incomplete for supporting conceptual design reasoning.
Heterogeneous types of variables and relations (non-numerical, integer, discrete and continuous).
Domain knowledge is available for detailed quantitative design; i.e. numerical analysis and optimisation routines, adaptation heuristics, and design cases. This knowledge, however, is incomplete and inadequate for generating conceptual designs. Generation of the best initial concept has traditionally been a manual process of trial and error guided by designers' experience. Global optimality is assessed by repetition of the conceptual design cycle and extensive search of the design space. Efficiency can be improved by reduction of the number and length of design iterations and application of abstract and qualitative domain knowledge to direct search into optimal subspaces. Three significant difficulties arise for the class of design problems considered:
Conceptual designing is an iterative process of progressive refinement. Design parameter types are inherently related to design decisions at specific levels of detail and phases in the design reasoning process.
Design problems cannot be decomposed into subproblems that can be solved independently due to strong component interaction; design problems have to be solved globally.
Strong component interaction and diversity of design parameter types both do not allow consistent transformations between quantitative and qualitative representations in an iterative process.