Bayesian inference as a tool for improving estimates of effective elastic parameters of wood

Abstract

A simple approach to the identification of geometrical and material uncertainties of wood is presented. This stochastic mechanics problem combines classical micromechanics, computational homogenization and experimental measurements with Bayesian inference to estimate the model parameters including the characteristics of errors in macroscopic elastic properties of wood caused by randomness of microstructural details on the one hand and the experimental errors on the other hand. The former source of uncertainty includes, for example, variability in microfibril angle and growth ring density. Even such limiting consideration of random input illustrates the need for combined computational and experimental approach in a reliable prediction of the desired material properties. Tying the two approaches in the framework of Bayesian statistical method proves useful when addressing their limitations and as such giving better notion on the credibility of the prediction. This is demonstrated here on one particular example of spruce wood.