Supporting concept synthesis by use of genetic algorithms

Evolutionary algorithms are common in parameter optimisation of a given solution. Conceptual design, however, initially involves finding a principle solution, and a bad principle choice can hardly be compensated by even the best detail design and parameter optimisation. This article therefore describes an approach to apply a genetic algorithm for concept synthesis by modelling principle choices as well as their parameter values and treating both as optimisation variables. The results are three alternative genome representations (of which a tree genome finally was chosen), algorithms for crossover and mutation operating on the tree genome and an evaluation strategy including weighted sum assessment over multiple criteria (conjunctive selection) as well as mechanisms that favour (disjunctive selection) or penalise (disjunctive elimination) concepts performing well or bad in one criterion respectively. The first task was to develop the method, which is also the focus of this article. Testing it in case studies in order to assess the effectiveness and practical applicability remains as a subsequent task. It can though be concluded that the presented approach fulfils its purpose, as it is capable of modelling the choice of the right principle solution during conceptual design.