Efficient Global Optimization (EGO) for Multi-Objective Problem and Data Mining

Abstract

In this study, a surrogate model is applied to multi-objective aerodynamic optimization design. For the balanced exploration and exploitation with the surrogate model, objective functions are converted to the Expected Improvements (EI) and these values are directly used as fitness values in the multi-objective optimization. Among the non-dominated solutions about EIs, additional sample points for the update of the Kriging model are selected. The present method is applied to a transonic airfoil design. In order to obtain the information about design space, two data mining techniques are applied to design results. One is analysis of variance (ANOVA) and the other is self-organizing map (SOM).