Probabilistic Sensitivity Studies of Homogenized Elastic Properties Predicted Using 3D Textile Mesoscale Models

The objective of this research was to attain a better understanding of the relationship between mesoscale model input parameters, such as tow and resin properties, on the homogenized elastic properties of a 3D textile composite representative volume element. This study was performed in support of a larger vision to develop a comprehensive and computationally tractable process-to-performance composites modeling framework. The study applied an approach for mesoscale model homogenization via the independent mesh method and periodic cluster method. The sensitivity analysis was performed using a Gaussian process response surface surrogate model approach. The results indicated that uncertainty in the model predictions could be reduced by gathering more data on the tow XY Poisson’s ratio and the tow fiber volume fraction for the warp and fill tows in the model. Additionally, the analysis indicated that spatial variability in the tow fiber volume fraction may be an important phenomenon that is missing from the mesoscale models. The value of these sensitivity analyses will be more fully realized when macroscale analyses on full-scale component and sub-component models are performed with the homogenized textile properties, which will help to elucidate the relationship between the textile architecture and the part/component response.