A neural network based approach for product form design

Abstract

A neural network based approach for product design is addressed in this article. Computer modeling, fuzzy set theory and semantic difference method are applied to set up an experiment. The experimental results are analyzed by applied back-propagation neural network, which establish the relationships between product–form parameters and adjective image words. A database for the connections among the design elements, product images and shape generation rules was constructed. A computer-aided system for product–form design was then developed based on this database. With the aid of this design system, a designer can generate 3D models of any product with different images by providing basic design elements and shape generation rules. Simultaneously, a rendered 3D model of the designed product and its images are also presented by this system. Therefore, changing the configuration parameter(s) until the product shape is acceptable can modify the image of a product. In this manner, the designed product can fit more closely to the consumers' desire. Chair design is taken as a case study; but this method can be used to develop other products.