Kriging: Methods and applications

Abstract

In this chapter we present Kriging-also known as a Gaussian process (GP) model-which is a relatively simple metamodel-or emulator or surrogate-of the corresponding complex simulation model. To select the input combinations to be simulated, we use Latin hypercube sampling (LHS); these combinations may have uniform and non-uniform distributions. Besides deterministic simulation we discuss random-or stochastic-simulation, which requires adjusting the design and analysis. We discuss sensitivity analysis of simulation models, using "functional analysis of variance" (FANOVA)-also known as Sobol sensitivity indices. Finally, we discuss optimization of the simulated system, including "robust" optimization.