Uncertainty quantification of wood floor system modal frequencies resulting from variable densities in members

Low-frequency vibration in a multi-story wood building is of concern because failure in mitigation of such vibration may lead to occupant discomfort. Ultimately, it may cause damage to their health. Variability in material properties of the members of a wooden floor system significantly influences vibration response characteristics such as eigenfrequencies and mode shapes. In this study, we investigate the uncertainty in low-frequency characteristics of a selected wood floor system, resulting from uncertainty in the members’ material properties. We use a surrogate model developed using polynomial chaos expansion, based on a few training samples computed using a finite element model of an experimental-scale structure comprised of seven load-bearing joists of spruce. We estimate the hyperparameters needed for accurate surrogate model development and, with these, we predict probability distributions for each of the floor-system modal frequencies. Also, we perform local and global sensitivity analyses to assess how material property variation affects the modal frequencies. Results confirm the accuracy and efficiency of the developed surrogate models in prediction of the modal frequencies, resulting from uncertainty in wood material properties. The accuracy of the results is comparable to those based on more expensive Monte Carlo simulation