Title: Multi-objective optimisation of split-plot designs
Authors: Kalliopi Mylona - King's College London (United Kingdom) [presenting]
Matteo Borrotti - University of Milan-Bicocca (Italy)
Francesco Sambo - Verizon Connect Research (Italy)
Abstract: Scientists can now address scientific issues of increasing complexity thanks to modern experiments. Often, factors in experiments have levels that are more difficult to set than others and in these cases, using a split-plot design offers a solution. Numerous approaches to finding optimal designs focus on maximising a specific criterion. To tackle the drawbacks of one-objective optimisation, multi-criteria techniques have been developed; however, they mostly concentrate on measuring the precision of fixed factor effects, ignoring the estimation of the variance components in split-plot experiments. To achieve pure-error estimates of the variance components, the Multi-Stratum Two-Phase Local Search (MS-TPLS) algorithm for multi-objective optimisation of experimental designs is expanded. The best Pareto front and associated designs, for motivating examples, are evaluated against other designs from the literature. According to the findings, the designs derived from the Pareto front are strong candidates for solutions based on the various objectives.