Kriging-based design of sequential experiment via best neighborhoods for small sample optimization

Surrogate model-assisted design of experiment has been fully developed in recent years, providing great help for small sample experimental optimization. A best neighborhood method for sequential experimental design of surrogate models is proposed in this paper. First, we use the Synthetic Minority Oversampling Technique for data enhancement on the small amount of data obtained in the experiment, Kriging method is used for surrogate modeling; then, Kriging prediction error and expectation improvement algorithm are used to set new sample point for a best neighborhood domain; finally, the best sampling points are selected through the genetic algorithm in the best neighborhood domain for sequential experiments. The proposed method is demonstrated with a numerical example and used for a validation case of low-speed flow analysis of naca0012 airfoil. The kriging-based design of sequential experiment via best neighborhoods can be useful for further experiment in any other engineering problems that need sequential sampling.