Consideration of the non-linear behaviour of joints for efficient design of complex timber structures

Licentiate thesis, 2023

The re-emergence of timber as a construction material, facilitated by the development of modern engineered wood products, has enabled the creation
of modern and complex timber structures. A detailed description of the load-displacement curves of timber joints is required. However, current design rules based on a semi-probabilistic design concept, do not capture the interactions among the components (system effects). In this regard, a reliability approach is more appropriate for evaluating the system effects and safety of the entire system. However, the estimation of the reliability of complex structural systems can be demanding in terms of computational effort and accuracy.
In this thesis, the impact of the non-linear behaviour of joints has been assessed by performing a parametric study on a simple but representative structure (beam supported by rotational springs at the ends). The uncertainties arising from the load, material characteristics, and the mechanical behaviour of joints were considered in the analysis. Different methods to estimate the reliability of a structure, such as Crude Monte Carlo, Important sampling, and metamodel-based techniques, were compared in terms of computational time and precision. Results showed that Crude Monte Carlo was the most adequate compared to other methods since the probability of failure was estimated with good accuracy and relatively small computational time. It was found that the probability of failure of the structure is sensitive to the elastic stiffness and ductility of the joints and their variability. However, a simple elastic-perfectly plastic model for joints is not representative of all the load-displacement curves of the joints. For this reason, regression models were applied to relevant load-displacement curves from the literature and their suitability for approximating diverse load-displacement curves was evaluated. The Richard-Abbott model was found to be the most appropriate for approximating the load-displacement shapes of joints and the associated variability.
This work provides the basis for a reliability-based approach for the analysis of structural systems, where the impact of joints on the performance of a structure is assessed and quantified. This will enable a more rational design of modern timber structures.