Hybrid Latin Hypercube Designs

Design of Experiment is becoming a common practice in solving many of engineering problems. In particular, sampling optimization is believed to be a major step in design optimization. Among the sampling methodologies, Latin Hypercube Design, whose efficiency was proven for wide range of applications, the paper addresses its optimization by proposing a multi-objective optimization criteria, which will be referred to as Hybrid Latin Hypercube Design (HLHD). Latin Hypercube has two major criteria, namely maximizing the minimum inter-sample distance and minimizing the inter-sample correlation. Unfortunately, the designs obtained by these two criteria can be entirely different and the optimization of one criterion does not necessarily lead to the optimization of the other criterion. However, the paper seeks the intersection area where the two objectives can meet.