A surrogate model to describe uncertainties in wood floor modal frequencies
Abstract
Wooden multi-story buildings have increased their market share in Europe over the last decades, mainly due to their relatively small carbon footprint and rapid onsite construction. Compared to more conventional heavy concrete buildings, wooden buildings are more sensitive to dynamic loads at low frequencies. Moreover, the dynamic response is generally more difficult to predict. This is, for example, due to the variability in the material properties of wood. Since wood is a natural material, local variations in performance occur, even if selected members are taken from the same batch. Also, the cut and treatment of individual members have an impact, and material properties can vary with time due to changing moisture content and aging effects. In this study, we investigate the effect of variability in material properties of joists on the dynamic modal characteristics of wood floors with the help of surrogate models, constructed using polynomial chaos expansion (PCE). These surrogate models are based on computations that use a dynamic reduced-order finite-element (FE) model, calibrated to produce modal outputs consistent with floor vibration experiments. Uncertainty in modal frequencies of the wood floor systems based on the PCE surrogate models is consistent with that found using the FE model in Monte Carlo simulations, which serve to define the “truth” model. However, PCE surrogate models offer significant reduction in computation time compared to FE models, when uncertainty in dynamic behavior needs to be quantified.
