A Novel Progressive Response Surface Method for High-order polynomial Metamodel

Abstract

The response surface method (RSM) model as a general metamodel has already widely been applied to many practical problems. However, the majority of studies only focus on the 2nd-order terms RSM model instead of how to determine the high-order terms. This paper introduced a novel progressive response surface method (NPRSM) combining optimal Latin hypercube design (OLHD), moving least square method (MLSM) and statistical test. This NPRSM model can decide the high-order terms by progressive procedure. Then three numerical function can be proposed to validate the accuracy and the approximated performance of the NPRSM model.