Efficient quantification of composite spatial variability: A multiscale framework that captures intercorrelation

Composite structures suffer from material imperfections. Non-deterministic models at the micro- and mesoscale propagate this spatial variability. However, they become impractical when the structure size increases. This paper proposes a numerically efficient multiscale methodology that links structural behaviour with the spatial variability of material imperfections on smaller scales. Fibre strength variability is accounted for through a fibre break model. A mesoscale model considers fibre volume fraction and fibre misalignment variability using random fields. Measurements provide probabilistic data for these imperfections. Subsequent homogenisation results in intercorrelated material properties on the structural macroscale that are modelled effectively with vine copulas. The methodology is verified by predicting the failure load of a coupon model. Predictions are very similar to those obtained by directly modelling spatial variability on the structural scale. •Propagation of uncertainty of unidirectional carbon-fibre-reinforced composite plies.•Stochastic fibre misalignment, volume fraction and fibre strength from measurements.•Intercorrelated finite element homogenised properties using vine copula models.•Efficient multiscale methodology accurately predicts unidirectional failure load.•Stochastic fibre misalignment and volume fraction reduce tensile strength considerably.