The effect of random microstructure morphology on the bending behavior of composite plate structures

The uncertain behavior of composite materials can be attributed to their random microstructure. Uncertainty propagation from micro- to macroscale is usually simulated using homogenization methods where boundary conditions (BCs), such as tension and shear, are applied to determine effective mechanical properties. This work focuses on plate structures, whose properties cannot be accurately estimated by classical homogenization schemes. Instead, they are described using the ABD stiffness matrix defined by classical laminate theory (CLT), and effective stiffness is obtained by application of bending/torsional BCs. The ABD-matrix homogenization method is applied, along with the moving window technique, for the computation of random fields of bending properties based on CT images of a real short fiber composite. Random fields describing bending and axial/shear stiffness are extracted, and their statistics are compared. Appropriate random field models are selected based on the marginal distributions and autocorrelation functions of the microstructure-derived random fields, while cross-correlations between different stiffness components are also considered. Finally, these random field models are employed for the stochastic finite element analysis of composite plate structures, whose response variability is investigated in detail. Useful conclusions are derived regarding the effect of several parameters on the response variability, including model size, scale factor and cross-correlations.