Neuro-fuzzy case-based design: An application in structural design

Abstract

This chapter presents a design approach and a design tool based on human analogical reasoning. This design tool provides solutions and it is used to design the formworks for constructing slabs. Through a case-based design process, past solutions are compared with the current design case, and a subset of them is retrieved according to a defined similarity measure. In this tool, the retrieval process is performed on the basis of a competitive neural network, which is submitted to unsupervised training by using the existing design solutions. The design case is represented in terms of sets of design parameters and associated fuzzy preferences. In engineering design problems whose solutions cannot be adapted in order to meet new requirements, the adaptation process is substituted by another approach that evaluates the retrieved design solutions according to the aggregation of the fuzzy preferences assigned to the current design problem. Therefore, the highly evaluated solutions may be manually adapted and modified by the designer based on his/her creativity and experience. In engineering domains, such as structural engineering design, which cannot be modeled computationally due to many different underlying disciplines, the designer's personal capabilities may be augmented by this design tool as it substantially assists decision-making.