Proposal of the method for updating of the response surface approximation and identifying information on correlation among design parameters

Abstract

The response surface approximation with uncertainty can be updated using a new method based on Bayes theory for the purpose of estimating the reliability more accurately in combination with data of numerical experiments. Furthermore, it is indicated that in the case that some response surface approximations for design space were stored and many sampling data from these approximations can be generated easily, the information of the design variables such as interaction among variables can be extracted and identified by applying various response surface models with many interaction terms actively.