Sandwich beams valued for its excellent mechanical performance and energy absorption characteristics. In this context, the paper introduce a new statistical approach based on a principal component analysis (PCA) using FatoMineR package for optimizing the design of waste core based-sandwich structures on the basis of an experimental database. Indeed, GFRP waste, sourced from a boat-building factory, was used to create several samples with polyester resin and evaluated for their physical properties and then assessed for thermal conductivity using a CT scale, with the core placed between two identical samples. The main outcomes provided by R software statistically highlight and improve the geometric features to maximize the features of the studied members.