1995 Volume 8 Issue 9 Pages 458-465
The case-based approach applied to the design problem may be promising alternative to the knowledge-based one, since it is free from the knowledge acquisition from human designers. However the generated design candidates are frequently restricted within the range of the past cases from the conventional case-based reasoning. In this article, we discuss an application of the genetic algorithm to establish the structural case-base and to generate design candidates. Each structure is represented as an individual and its fitness is evaluated in terms of the frequency in the past cases. The case-base population is formed through the genetic operations. The genetic operation is also applied to the case-base for the candidate generation. The similarity between the design specification and that in the case-base is used as the fitness at this stage. The candidates generated by the proposed operations reflect the past cases but are partially free from them. This gives diversity of the design candidates to be considered by human designer.