Mesoscale simulation of woven composite design decisions

TL;DR

This paper uses mesoscale finite element simulations and global sensitivity analysis to explore how material design choices affect the thermophysical properties of woven composites, aiding accelerated materials development.

Contribution

It introduces a comprehensive sensitivity analysis framework linking mesoscale design parameters to composite properties, integrating surrogate modeling and Sobol' indices.

Findings

Both constituent properties and mesoscale geometry significantly influence composite properties.

The surrogate model effectively captures parameter-property relationships.

Sensitivity analysis identifies key parameters impacting material performance.

Abstract

Characterizing the connection between material design decisions/parameters and their effective properties allows for accelerated materials development and optimization. We present a global sensitivity analysis of woven composite thermophysical properties, including density, volume fraction, thermal conductivity, specific heat, moduli, permeability, and tortuosity, predicted using mesoscale finite element simulations. The mesoscale simulations use microscale approximations for the tow and matrix phases. We performed Latin hypercube sampling of viable input parameter ranges, and the resulting effective property distributions are analyzed using a surrogate model to determine the correlations between material parameters and responses, interactions between properties, and finally Sobol' indices and sensitivities. We demonstrate that both constituent physical properties and the mesoscale geometry strongly influence the composite material properties.