Comparison Between Different Types of Latin Hypercube Designs Applied to Kriging Models

Abstract

Numerous scientific fields employ Kriging models (KMs) to study how physical systems behave. The computer code's response is interpreted by the KM as Gaussian process. A well-designed is an essential part of the KMs. Our study involves the design and analysis that is used in constructing KMs for computer experiments. Evaluating how design affects prediction accuracy is what we are interested in. Latin hypercube designs (LHDs)  are commonly used for building KMs. We investigate many types of LHDs and assess their performances. We built many KMs based on different types of LHDs and the performance of each model is compared. The comparison is performed through some measures that test the KM performance. We apply KMs to the Robot Arm function as a real computer code.