Knowledge Discovery in Dataset Generated by Topology Optimization

In this paper, we propose a new framework to obtain knowledges for appropriate formulations in topology optimization. The basic idea of this framework is incorporating Knowledge Discovery in Databases (KDD) and topology optimization. That is, we construct a database that records various material distributions generated by topology optimization for various structural design problems, including the varieties of design domain shapes, boundary conditions, and performance indicator functions, and then, we find appropriate material distributions in the database for a given design domain, boundary condition, and performance indicator functions. Furthermore, the proposed framework is designed such that we can easily change the performance indicator functions used to evaluate the material distributions. By doing so, we can easily test various indicator functions and obtain knowledges for appropriate setting of indicator functions, i.e., formulation. In contrast to conventional topology optimization methods which find optimized material distributions under given formulations, the proposed framework aims to find appropriate formulations themselves. We present a simple prototype system based on the proposed framework and discuss the potential of the proposed framework.