Shape-Generation Method Using Curvature Entropy.

Abstract

In curve design, it is important to grasp and control the whole characteristic of shape that emerges from combinations of shape elements. However, it is difficult that only conventional shape-information expresses the whole characteristic of shape. Therefore, a new shape-information that could express the whole characteristic of shape and a design-support system using it as a design guideline are desired.In past research, the authors have defined the statistical amount of shape (SAS) from information entropy and expectation value. Principal component analysis of simple curve profiles, such as geometric figures, was carried out using conventional shape information and SAS. Curvature entropy and curvature expectation value were included in the independent main component, and the relationship between SAS and various types of recognition information in the simple curve profile was analyzed. The correlation between curvature entropy and “complexity”recognition information was about 0.7. Curvature entropy is thus unlike conventional shape information. It can also express “complexity”characteristics in a simple curve profile.In this study, we described shape-generation method using curvature entropy and genetic algorithm, and applied this method for the design of the automobile side view. As a result, it was confirmed that curvature entropy could express the whole characteristic of shape in a curve profile at the product level, and the possibility of the design-support system using curvature entropy was shown.