Comparison of unit cell-based computational methods for predicting the strength of wood

Wood, as a naturally-grown material, exhibits a highly anisotropic and inhomogeneous material structure, with a complex wood fibre distribution influenced by randomly occurring knots. Thus, for the prediction of effective strength properties of wood, advanced computational tools are required, which are able to predict as well as consider multidimensional strength information at different scales of observation. Within this work, three such computational methods will be presented: an extended finite element approach able to describe strong strain-softening and, thus, reproduce brittle failure modes accurately; a newly-developed limit analysis approach, exclusively describing ductile failure; and an elastic limit approach based on continuum micromechanics. All three methods are applied to earlywood and latewood unit cells and to clear wood, finally yielding effective failure surfaces for a range of multidimensional stress states. These failure surfaces are compared with each other and with experimental results from biaxial tests. Based on these comparisons, the strengths and weaknesses of the three computational methods are discussed, and their applicability to wood is evaluated. The extended finite element method is a powerful technique that allows for a very realistic description of strength-governing processes. Nevertheless, its complexity and high computational effort prevent widespread use in the engineering field. The plastic limit analysis and elastic limit approaches, however, show good predictive performance compared with the extended finite element method, coupled with excellent efficiency and stability. In this study it is found that together, the latter two approaches are able to enclose the experimentally-obtained failure regions for clear wood almost perfectly, while also delivering new insights with respect to the ductile failure potential of wood. The conclusion can be drawn that there exist promising computational methods that are capable of delivering reliable multidimensional strength information for wood and, subsequently, will enable effective strength predictions for wooden boards and wood-based products. Finally, this work is intended as a contribution to performance-based optimisation of wooden structures, a necessity for wood to become competitive with respect to other building materials.