The Global Stiffness of a Multi-Storey Timber Building

Abstract

This thesis investigates the global stiffness of cross-laminated timber (CLT) buildings. A case study is conducted on the 10-storey Urban Woods building in Delft, in which the lateral load-resisting system consists of façade bracing supported by a central CLT core. The study evaluates the accuracy of early-stage analytical and numerical models in predicting global stiffness and examines the influence of uncertainty in the slip modulus of dowel-type connections. The objective is to improve the understanding of the applicability of early-stage modelling approaches for estimating the global stiffness of timber structures, and to quantify the impact of uncertainties in connection stiffness on the global structural response.

Assessing the global stiffness of timber buildings is critical, as their lower stiffness and self-weight compared to conventional concrete, masonry, and steel structures make them more susceptible to global deformations. The global stiffness is governed by both the configuration of the structural system and the stiffness of its connections, particularly within the lateral load-resisting system.
During the preliminary design phase, engineers must estimate global stiffness and the associated horizontal deformations of a proposed structure. However, due to the relatively limited number of timber high-rise buildings and the associated lack of practical experience, these estimations remain uncertain and need to be supported by complex numerical models. In addition, recent studies have reported discrepancies between experimentally obtained slip moduli and analytical predictions according to Eurocode 5 for dowel-type connections. As these fasteners are widely used in timber lateral load-resisting systems, this introduces additional uncertainty in predicting global structural deformations.

The translational stiffness of the connections was determined using the analytical serviceability approach prescribed in Eurocode 5. These stiffness values were implemented in both analytical models based on the method of Virtual Work and numerical models developed in RFEM. The models were subjected to self-weight and wind loads according to serviceability limit state combinations. To assess the influence of connection stiffness, the translational stiffness values were varied, and the resulting maximum horizontal deflection was used as the primary response parameter.

The results show that analytical models based on the method of Virtual Work systematically overestimate the global stiffness of timber structures by approximately 10–40% compared to numerical models developed in RFEM, primarily due to simplifying the structure. A key difference between the modelling approaches lies in their sensitivity to changes in connection stiffness. For moderate reductions in connection stiffness (up to approximately 50–60%), the overestimation remains relatively constant. However, beyond this threshold, a significant divergence emerges: the numerical models exhibit a non-linear decrease in global stiffness, whereas the analytical models predict a linear reduction. For the global structural model, a reduction of 60% in connection stiffness results in an average decrease in global stiffness of approximately 20%. These findings are supported by literature, which reports that analytical estimates of connection stiffness based on Eurocode 5 can overestimate stiffness by up to approximately 60%, indicating that such reductions are feasible in practice. Therefore, it is advised to increase the estimation of the horizontal deflection by up to 20%, if the lateral load-resisting system is dependent on the stiffness dowel-type connections.

The results further indicate that the accuracy of analytical and numerical models depend on the structural complexity, with better agreement observed for simpler configurations. In addition, the importance of form stability in the lateral load-resisting system is highlighted: structures with a higher degree of form stability are less sensitive to the stiffness of connections. For systems made out of simple and form stable configurations an analytical model can provide reliable estimations of the global stiffness. However, numerical models tend to perform better and are therefore preferred for the analysis of more complex and non-form stable structures.