Abstract

     Robustness research has become popular, however very little is known on its explicit quantification. This paper summarises a quantification method previously published by the main author and proceeds in demonstrating its step-by-step application with a case study tall timber building. A hypothetical 15-storey post-and-beam timber building with a central core is designed for normal loads, and four improved options are designed to account for abnormal loads in order to increase the building’s robustness. A detailed, nonlinear, dynamic Finite Element model is set up in Abaqus® to model three ground floor column removal scenarios, and a Random Forest classifier is set up to propagate uncertainties, to efficiently estimate the probability of certain collapse classes occurring, and to calculate the importance of each input parameter. The results show how design improvements at the whole building scale (e.g., strong floors) have a higher impact on robustness performance than just improving the strength and ductility of some selected connections, although these results are exclusive to the building studied. The case study reinforces the importance of a sound conceptual design for achieving robustness in tall timber buildings.

Introduction

     Our profession is seeing a paradigm shift towards buildings of lower carbon footprint: tall timber buildings are a fine example of this shift. Any departure from the “familiar waters” of the common structural typologies runs a higher risk of not having identified or anticipated certain structural behaviours: similar to the scaling issues that led to the partial collapse of the Ronan Point in London in 1968 [1], we must understand how timber buildings scale to the new heights constructed in the last 10 years, particularly regarding their disproportionate collapse behaviour.

     Structural robustness, or disproportionate collapse resistance, is the ability of a structure to withstand damage without disproportionate further consequences, and it is an important and yet not so widely understood quality of our building stock. While a lot of work has been put in understanding structural robustness at a qualitative level and regarding concrete and steel buildings, little is known on the robustness of timber buildings and even less on how to specifically quantify how robust is a building, and whether this is enough or not. Initial studies focused on medium-rise CLT buildings have identified connection ductility as a key requirement to enable catenary action in beams and floors, an efficient way to redistribute loads in case of damage.

Conclusion

     The work presented in this paper is an advanced robustness quantification of a tall timber building. The quantification procedure was only made possible by some advanced modelling techniques. In particular:

• We designed an imaginary 15-storey CLT-core, post-and-beam tall timber building and four additional improved versions of it, all according to the Swiss building codes. Particular attention was put into calculating the stiffness and strength of all connections in all degrees of freedom.

• The buildings were then analysed using detailed nonlinear dynamic Finite Element models in three ground floor column removal scenarios to calculate the resulting robustness indices and to compare their disproportionate collapse performance.

• The robustness indices were calculated by using a Random Forest classifier. Model enrichment with the Synthetic Minority Oversampling TEchnique (SMOTE) algorithm were used to improve the classifier performance.