Technological advances in engineered timber structural elements have led to an increase in the popularity of mid-rise and high-rise timber buildings. The governing design criterion of such buildings is very often serviceability under wind-induced vibration. Accurate finite element models for predicting their modal properties (i.e. natural frequencies and mode shapes) are crucial for designing buildings that satisfy the current serviceability criteria. It is a challenge for structural engineers to decide what to include in the structural modelling. This is because elements that are typically considered non-structural (partition walls, plasterboards, screed, façade, etc.) have been shown to act structurally and can significantly influence the modal properties of timber buildings. This thesis presents three case studies of timber buildings, where finite element models were developed. Using experimental modal properties, the models were updated to learn about the as-built stiffness characteristics of the analysed buildings. A probabilistic framework based on Bayesian inference was applied to account for the uncertainties. A polynomial chaos expansion was used to develop a surrogate model allowing for a more time-efficient conducting of stochastic analyses such as uncertainty quantification, sensitivity analysis, and Markov chain Monte Carlo posterior sampling. Deterministic and probabilistic approaches were compared and assessed for their effectiveness.
|
