Hierarchical modeling of structural timber material properties by means of Bayesian Probabilistic Networks

In the pursuit of modeling the performance of timber structures the probabilistic representation of the variability of timber material properties is an issue of special interest. Material properties can be represented by random variables and the statistical characteristics of these variables can be described by distribution models together with the corresponding parameters which are calibrated based on data taken from standard test samples and grading machine measurements. In the present paper special emphasis is directed on how to represent multi-scale spatial variability of timber material properties. Variability of timber material properties is considered at different levels and, subsequently, brought together into one consistent hierarchical model by means of a Bayesian probabilistic network. The hierarchical model is used to determine the influence of the origins and cross-sectional dimensions of the timber on the probability distribution of timber material properties.