Regression metamodeling in computer simulation — the state of the art

Abstract

This paper discusses the growing use of metamodels in analyzing manufacturing/service systems. Metamodels have tremendous appeal for practical purposes; they utilize the increasing power of PC-based simulations. This survey identifies the different experimental designs that have been applied as well as their benefits and limitations. A detailed explanation on how they are used is also provided, with citations to the specific cases that have been modeled. Specifically, the use of group screening, full and fractional designs, and polynomial decomposition, in developing regression metamodels are discussed. This exposition provides, in a nutshell, the research studies in the area and also offers practitioners an alternative approach to model complex problems. A framework on how to develop a regression metamodel is provided. Furthermore, an application of the framework to a maintenance float system is demonstrated.