Fast simulation of non-stationary wind velocity based on time-frequency interpolation

The wind non-stationary features have a great influence on structural response analysis. The spectral representation method widely used to generate the non-stationary excitation suffers from lower efficiency when the simulation points become large. The Cholesky decomposition is a critical step and also a time-consuming procedure. Although the existing interpolation method is conducted in frequency direction, the number of Cholesky decomposition is still large. In this study, the time-frequency interpolation scheme is proposed to conduct two rounds of spline interpolation along frequency direction and time direction respectively. Then the quartic polynomial transform is developed to transform the uniform distribution of frequency into non-uniform distribution. The number of interpolation points is defined by error and computational time. Numerical example involving the simulation of fluctuating wind velocities along vertical direction shows that the proposed method reflects the target spectrum very well and maintains steadily high efficiency in the interpolation procedure of Cholesky decomposition. Comparing with existing interpolation method, more than 60% of computational time can be saved, and the accuracy of the two methods is close to each other. The time-frequency interpolation speeding up decomposition greatly is suitable for simulation of non-stationary wind velocity in large structures.