Adaptive Kriging Modeling Using Continuous and Multi-modal Characteristics Exploration

To build a highly accurate metamodel with high efficiency, an adaptive Kriging modeling method using the exploration of continuous and multi-modal characteristics was proposed. After obtaining an initial Kriging model, the leave-one-out cross-validation strategy was utilized to calculate the metamodel’s accuracy. Then the sample point with the maximum error was selected from the sample database according to the relative error criterion. A new sample point and its approximate response value were acquired quickly by using the Taylor Series expansion. With the above procedures repeated, the sample database and the Kriging model were updated in order to make the Kriging model achieve the required accuracy. The method was applied to two mathematical problems and an engineering problem. Results show that the method can get a higher accurate Kriging model with higher efficiency.