Spectrum Models for Nonstationary Extreme Winds

Nonstationary extreme winds are responsible for many structural damages. However, empirical models are not available for these winds, especially for fluctuations, mainly due to the difficulties in mathematical treatments. In this study, nonstationary wind characteristics are studied and analytic models are also proposed based on field measurements of wind speeds. Firstly, the discrete wavelet transform and kernel regression method are used to infer the time-varying mean and variance of the nonstationary extreme wind, respectively. Then, based on the estimated evolutionary power spectral density (EPSD), transient features of nonstationary winds are examined. Results show that spectral variations in nonstationary wind fluctuations including concerned downbursts and typhoons are relatively weak. This means that these nonstationary fluctuations can be modeled as uniformly modulated processes. Also, the validity of nonstationary wind spectrum models directly extended from current stationary wind spectra is evaluated. The study demonstrates that this extension is not appropriate. Furthermore, two analytical models are suggested to characterize nonstationary wind fluctuations, including a fully nonstationary process model and a simplified uniformly modulated process model. Both models have physical meaning and provide satisfactory fitting for the estimated EPSD. They will be helpful in the Monte Carlo simulation and structural dynamic analysis. An evaluation of these models is conducted based on the structural dynamic analysis using a series of tall buildings. Results show that these two models have good performance in structural response prediction.