부산대학교 도서관

Design concept generation is an important task to produce innovate products. While various methods have been proposed for supporting the task, their impacts depend on designer’s creativity and expertise. The objective of this research is to propose a computational methodology for supporting design concept generation with the integrated use of topology optimization and neural network. It is intended that the methodology facilitates a designer to generate various and promising design alternatives by topology optimization, and to find beneficial design concepts through classification of them by neural network. This paper reports experimental implementation of a neural-network-based concept extraction method toward the realization of such a methodology. Numerical examples of conceptual design of bridges demonstrate its fundamental capability in categorizing design alternatives, which are generated through topology optimization, and identifying the spot to define new design concept.