# Multiscale Uncertainty Quantification of Woven Composite Structures by Dual-Correlation Sampling for Stochastic Mechanical Behavior

## TL;DR

This paper introduces a new method to better understand and predict the uncertain mechanical behavior of woven composite materials by accounting for correlations at multiple scales.

## Contribution

A novel dual-correlation sampling approach is proposed for multiscale uncertainty quantification in woven composites with physical interpretability.

## Key findings

- The dual-correlation sampling method accurately captures spatial autocorrelation and cross-correlation in material properties.
- Experimental validation confirmed accurate predictions of probabilistic mechanical responses and damage morphology.
- Traditional methods fail to represent spatial correlations, leading to inaccurate structural behavior predictions.

## Abstract

Woven composite structures are inherently influenced by uncertainties across multiple scales, ranging from constituent material properties to mesoscale geometric variations. These uncertainties give rise to both spatial autocorrelation and cross-correlation among material parameters, resulting in stochastic strength performance and damage morphology at the macroscopic structural level. This study established a comprehensive multiscale uncertainty quantification framework to systematically propagate uncertainties from the microscale to the macroscale. A novel dual-correlation sampling approach, based on multivariate random field (MRF) theory, was proposed to simultaneously capture spatial autocorrelation and cross-correlation with clear physical interpretability. This method enabled a realistic representation of both inter-specimen variability and intra-specimen heterogeneity of material properties. Experimental validation via in-plane tensile tests demonstrated that the proposed approach accurately predicts not only probabilistic mechanical responses but also discrete damage morphology in woven composite structures. In contrast, traditional independent sampling methods exhibited inherent limitations in representing spatially distributed correlations of material properties, leading to inaccurate predictions of stochastic structural behavior. The findings offered valuable insights into structural reliability assessment and risk management in engineering applications.