Global Feature based IREC (Applied to Curved Shapes in Industrial Design)

Abstract

In this paper, G-IREC is proposed. G-IREC is an extended IREC, which pays special attention to global design features which are in the form of curved shapes. IREC (Interactive Reduct Evolutional Computation) is a method to evolve designs based on users' personal preferences. The method works through an interaction between the user and a computer system. The computer system with IREC generates design samples consisting of random attributes and the user evaluates and scores each samples depending on his/her psychological preferences. The system estimates the design attributes that the user pays more attention to (favored features) with Reduct in Rough Sets Theory and reflects it to generate new design samples. This interaction continues until the samples converge to a satisfactory design. So far design parameters such as coordinate of nodes for spline curve have been regarded as design attributes. However, design attributes consist of not only detailed local parameters but also global features such as convexity, softness etc. In earlier design process, designers first drew a rough sketch to determine the outline of a design, and designed the details later. Therefore it is important to introduce global features to accelerate the convergence and to increase user friendliness. We develop G-IREC (Global feature based IREC) which allows to introduce global features in IREC. G-IREC (Global feature based IREC), which introduces global features into IREC, is applied to an automobile side board shape design. The effectiveness of this method is demonstrated by comparing the computed results with the experimental ones.