The influence of experimental design method on design space exploration

Abstract

Simulation-based product design optimization includes processes of sampling using design of experiment (DOE), numerical simulation and construction of optimization model. The suitable DOE method is the one that can generate a set of sample points that better expresses the characteristics of the design space, making the design optimization process more accurate and efficient. We perform design space exploration and focus on the influence of DOE method on data mining. Firstly, different DOE methods are used to generate sample points. Secondly, design space exploration is carried out using rough set theory. Then analyze the effect of space exploration from the aspect of accuracy, the degree of space reduction, the number of sample points required and the like, after which an appropriate experimental design method for the problem is determined. The results show that in the problem of data mining for design optimization, uniform design is superior to Latin hypercube design, which lays the foundation for further application of uniform design in product design optimization.