Numerical simulation of non-Gaussian wind load

Abstract

Compared with Gaussian wind loads, there is a higher probability of strong suction fluctuations occurrence for non-Gaussian wind pressures. These instantaneous and intermittent fluctuations are the initial cause of local damage to roof structures, particularly at the edges and corners of long-span roofs. Thus, comparative errors would occur if a Gaussian model is used to describe a non-Gaussian wind load, and structural security would not be guaranteed. This paper presents a simplified method based on the inverse fast Fourier transform (IFFT), in which the amplitude spectrum is established via a target power spectrum. Also, the phase spectrum is constructed by introducing the exponential peak generation (EPG) model. Finally, a random process can be generated via IFFT that meets the specified power spectral density (PSD), skewness and kurtosis. In contrast to a wind tunnel experiment, this method can avoid the coupled relation between the non-Gaussian and the power spectrum characteristics, and lead to the desired computational efficiency. Its fitting accuracy is not affected by phase spectrum. Moreover, the fitting precision of the kurtosis and PSD parameters can be guaranteed. In a few cases, the fitting precision of the skewness parameter is fairly poor, but kurtosis is more important than skewness in the description of the non-Gaussian characteristics. Above all, this algorithm is simple and stable and would be an effective method to simulate a non-Gaussian signal.