Prototype of a User Friendly Shape Design System with Sense Learning Mechanism

Abstract

A user friendly shape design system which fits an user's sense has been developed. The system has dials as operation devices. The dials can be defined by words which express user's ideas. The system uses a neural network to convert user's ideas which are input by the dials to geometric parameters. Using the neural network, relations between user's ideas and geometric parameters can be defined while the relations are difficult to define mathematically or nonlinear. Learning data to define each word's transformation function can be generated by an user's selection of appropriate sets of source and destination shape samples for an idea expressed by the word. This learning procedure enables an user to make neural network parameters reflected the user's sense. The system was evaluated by toy plane design experiments.