A Unified Probabilistic Approach for Predicting the Structural Response of Oriented Strandboard

McTigue, Anthony
Oriented strandboard (OSB) is a wood-based composite manufactured from elongated wood-strands coated in a resin binder that are oriented and arranged in layers, cross-laminated and hot pressed to form large panels. OSB is a highly complex material that requires a large number of parameters to fully define its internal structure and mechanical behaviour. The variability of the defining parameters further adds to the complexity. Current design and production practices rely on highly-simplified, deterministic methods where many of the defining parameters and their variability are omitted and replaced with high safety factors. Reliability methods offer significant potential for improved efficiency in the OSB industry. However, such methods require knowledge of the stochastic properties, mechanical behaviour and relationships between parameters as well as appropriate modelling and analysis tools. In this thesis, a new approach to predicting the mechanical behaviour and associated variability of OSB/3 panels using a stochastic finite element model is developed. As part of the work, a large-scale experimental programme was undertaken that included over 2,780 tests to evaluate 45 different physical and mechanical properties for commercial OSB/3 and single-layer OSB panels. This provided the necessary information to evaluate the stochastic properties, mechanical behaviour and correlations between parameters. The Anderson-Darling method was used to establish suitable probabilistic models to describe the variability of each parameter. The Pearson's correlation coefficient was used to describe relationships between the parameters. Using regression analysis, suitable mathematical models to accurately represent the orthotropic, non-linear mechanical behaviour of OSB subjected to tension, compression, bending and panel-shear loading were developed. New normalised non-linear constitutive relationships for OSB were implemented in 3-D finite element models. A stochastic analysis based on the Monte Carlo method accurately reproduced the variability of the defining parameters found in the experiments. This approach is suitable for implementation in reliability-based structural design.
Publisher DOI
Attribution-NonCommercial-NoDerivs 3.0 Ireland