Development of Visualization Model for Mechanical Design Knowledge and Estimation of Learning Support System on Self-Organizing Maps

摘要

This paper deals with the mechanical design support system which uses Self-Organizing Maps (SOM). As the wok cost on mechanical design is generally required to keep in low, the reusability of the past design cases is quite important Although expert engineers can search appropriate design parts from the past cases and can be reminded its usage or can estimate its usage, the support system is seemed to be useful and necessary for many novices to search such the reusable cases and to understand its usage. In this study, we report for developing a design support system based on the past design parts as the knowledge space. The support knowledge was figured by SOM being one of Neural-network application. Firefly, how to metric or describe the mechanical design knowledge in a Map was discussed. In the next step, a generation method of Map model for some practical demands in the design process, and an indication method of design knowledge required for training were described. In order to generate and verify the proposed SOM model in the support system, a spur gear design was considered as a case problem in a mechanical design space. The SOM model for the spur gear was developed and its simulation was carried out.