Abstract
Composite materials properties are affected by uncertainties that cannot be overlooked for accurate modelling predictions. In the present study, a novel implementation of statistical screening methods for sensitivity analysis on composites is proposed. The effect of uncertainties on the behaviour of the model is assessed rapidly and reliably. Despite their efficiency when models with several input factors are employed, screening approaches are rarely used in engineering. Two sampling strategies are explored, and the results for several case studies are shown and compared with statistical estimators from regression-based methods. It is shown that screening techniques manage to provide subsets of influential parameters for a variety of applications, including analytical and finite element models, with low computational cost.