Analyzing, modelling and simulating nonstationary non-Gaussian thunderstorm winds and their use for assessing the statistics of extreme wind pressure
Abstract
Thunderstorm wind speed varies rapidly, resulting in the time-varying mean wind speed and direction for the alongwind, and fluctuating alongwind, crosswind and vertical wind. The amplitude and frequency contents of fluctuating wind speeds are also time-varying, at least for the fluctuating alongwind and crosswind. This thesis proposed approaches for analyzing, modelling, and simulating thunderstorm winds with respect to their time- and directional-variation of the alongwind speed and direction, as well as the multidirectional fluctuating winds. The analysis and decomposition of thunderstorm winds were carried out using continuous harmonic wavelet transform and Fourier transform. The analysis results indicated that the power spectral density (PSD) functions of the fluctuating winds depend on the time-varying mean wind speed, and the fluctuating components can be treated approximately as Gaussian. Based on the time-frequency spectral analysis, the PSD functions and lagged coherence functions for the fluctuating winds in the alongwind and crosswind directions in the horizontal plane are developed; new empirical models for the time-varying mean wind speed in the alongwind direction and its corresponding direction are also established. This model was then extended to tri-directional thunderstorm winds, which involves determining the time-varying mean wind speed and direction by considering both azimuth and elevation angles and analyzing the spectra of high-frequency wind components in three orthogonal directions. The simulation of bi- or tri-directional thunderstorm winds based on the proposed models are illustrated and validated by numerical examples.
In addition to the development of stochastic models for thunderstorm winds, this thesis also develops record-based simulation methods for simulating multivariate thunderstorm winds; the methods make use of the instantaneous features of limited available data. The developed procedure is based on the framework of the iterative power and amplitude correction algorithm but with modifications and considers different time-frequency decomposition techniques. The modifications are aimed at increasing the variability of the sampled record components by randomizing the power spectral density functions of processes through a digital filter in the frequency domain, and improving the convergence by using a relaxation factor for the synchronized phase shift.
In addition to the analysis, modelling and simulation of the multi-component thunderstorm wind record, statistical analysis is carried out to investigate the effects of the time-varying mean wind speed and direction, and the nonstationary fluctuating components of thunderstorm winds on the extreme peak wind pressure and wind pressure coefficient for a low building by applying the equivalent-steady-gust model and the quasi-steady-vector model. The effect of the random orientation of wind direction on the estimated extreme wind pressure coefficient is also investigated. The gust front wind loading factors considering thunderstorm winds and atmospheric boundary layer synoptic winds are assessed based on probabilistic analysis.