In probabilistic composite mechanics, uncertainty modelling may be introduced at a constituent (micro-scale), ply (meso-scale) or component (macro-scale) level. Each of these approaches has particular advantages/limitations and appropriate fusing and benchmarking is desirable in order to improve confidence in probabilistic performance estimates of composite structures. In the present study, random variable based micro and macro-scale reliability analyses are critically compared through a limit state formulation based on the analytical stress tensor components of a rectangular simply supported orthotropic FRP composite plate and the Tsai–Hill failure criterion. The study aims to promote cross-fertilisation of alternative uncertainty modelling approaches in a multi-scale analysis framework. Propagation of uncertainty from micro to macro-scale, and the corresponding influence of changes in random variability on the reliability estimates is quantified. The importance of benchmarking experimentally-based probability distributions of mechanical properties through micro-scale modelling is illustrated, and the confidence that can be placed on reliability estimates is quantified through a series of numerical examples.
- polymer-matrix composites (PMCs)
- mechanical properties
- statistical properties/methods