A simple approach for incorporating soil-structure interaction and site uncertainty into floor response spectra

The knowledge of the floor response spectrum (FRS) is crucial for a sound evaluation of the seismic behaviour of non-structural components. However, accurately assessing the FRS is challenging due to both soil-structure interaction complexity and system uncertainties. This study examines the influence of site-specific soil uncertainties on the determination of FRS using a spectral stochastic finite element approach. By leveraging the random fields concept, the Karhunen-Loève expansion is applied to model a continuous non-Gaussian soil profile, integrating a polynomial chaos expansion to capture the variability of soil properties. This methodology is grounded on empirical data from a target site, enhancing its practical applicability, while employing an orthogonal autocorrelation function significantly improves modelling efficiency. An FRS analysis model that combines finite element method with random field characteristics, allowing for the derivation of a probabilistic FRS through spectrum-to-spectrum analysis, is developed. An illustrative example shows that the generated FRS aligns well with Monte Carlo simulation results. The findings indicate that uncertainties in soil parameters substantially affect the spectral mean, standard deviation, and peak position. This approach is particularly effective for generating probabilistic FRSs, especially in all those cases where a high level of confidence is needed.