SIMULATION OPTIMIZATION FOR COMPUTER MODELS WITH MULTIVARIATE OUTPUT

This paper explores the potential of Gaussian process based Metamodels for simulation optimization with multivariate outputs. Specifically we focus on Multivariate Gaussian process models established through separable and non-separable covariance structures. We discuss the advantages and drawbacks of each approach and their potential applicability in manufacturing systems. The advantageous features of the Multivariate Gaussian process models are then demonstrated in a case study for the optimization of manufacturing performance metrics.