Multiscale Stochastic Modeling of the Elastic Properties of Strand-Based Wood Composites

This paper introduces a novel modeling approach for wood composites using concepts of numerical homogenization employed in synthetic composites. It describes a multiscale model based on a unit cell that incorporates both the wood and resin phases for simulating structural composite lumber made of strands. In this approach, constant resin thickness and strand geometry, elastic properties of constituents, and perfect bonding between wood and resin are assumed. The multiscale modeling is composed of two steps. The first step estimates the effective elastic properties of a unit cell based on the numerical homogenization with periodic boundary conditions. The second step consists of a macroscopic finite element structural analysis of a beam (assembly of several unit cells) under three-point bending. Random distribution of strand orientation that may be encountered in an actual composite beam is introduced at this stage. Results indicate a significant influence of the resin. The first step of the approach provides an initial illustration when comparing effective properties of unit cells with different resin volume fractions and/or elastic properties. The resin decreases the Young's modulus of the unit cell in the fiber direction while strengthening the transverse and shear moduli depending on the contrast between the resin and wood properties. The final results obtained for the beam show that the bending modulus decreases with increasing resin content, which is a combined effect of the micro-mechanical phenomena. The key contribution of this paper to modeling wood composites is the addition of the resin phase to a numerical model and inclusion of random distribution of strand orientation. DOI: 10.1061/(ASCE)EM.1943-7889.0000381. (C) 2012 American Society of Civil Engineers.