A Weighted Sequential Sampling Method Considering Influences of Sample Qualities in Input and Output Parameter Spaces for Global Optimization

摘要

A new sampling method, namely weighted sequential sampling method, is introduced in this research to improve accuracy and efficiency of adaptive metamodeling considering influences of sample quality measures in both input and output parameter spaces. In this method, sample quality measures in input and output parameter spaces are associated with weighting factors. Values of these weighting factors are changed in sequential sampling considering the different levels of contributions of these sample quality measures in the input and output parameter spaces during the adaptive metamodeling process. Since quality of the metamodel developed through weighted sequential sampling is good in the whole design space, quality of global optimization can be improved through adaptive metamodeling based on weighted sequential sampling.