Simulation of fluctuating wind velocity at given position on moving high-speed train

The modified Karman wind spectrum was employed as the target spectrum. The order of the autoregressive model for the linear filtering method (LFM) was determined by using the Akaike information criterion (AIC), and the reliability of the simulation results was verified. The LFM and the harmonic superposition method (HSM) were used to simulate the fluctuating wind velocity at a given position on a moving high-speed train. The characteristic variables of fluctuating wind velocity, such as mean values, variances, amplitudes, phase frequencies and distributions calculated by the two methods were compared. The computational efficiencies of these two methods were analyzed. Analysis result shows that the power spectrum densities of fluctuating wind velocity obtained from the two methods fluctuate around the target power spectrum. The mean value of fluctuating wind velocity is approximately 0. Due to the existence of random number, there is a difference in the peak values of fluctuating wind velocity at certaintime, and the differences in the simulated power spectrums within low frequency area may exceed 50%. Under different yaw angles, the difference in the mean values of fluctuating wind velocity between the two methods is less than 2%, and the distribution rules are quite similar. When the train speed is 80 m·s^-1 and the average wind velocity at the height of 10 mabove ground level is 25 m·s^-1, the difference in the average peak values of fluctuating wind velocity between HSM and LFM is less than 1%. Both the two methods are suitable for simulating the fluctuating wind velocity at a given position on a moving high-speed train. With the increase of the frequency of fluctuating wind velocity, the amplitude of fluctuating wind velocity decreases and the phase fluctuates in the range of-πtoπ. The difference in the fluctuating wind velocity distributions in the range of-3 to 3 m·s^-1 is 0.48%. The point numbers of fluctuating wind velocity obtained by the two methods both correspond to the Gaussian distribution, and the maximum difference in the Gaussian distribution fitting coefficients is 3.15%. The variance of the fluctuating wind velocity obtained from LFM is 7.89% larger than that obtained from HSM, therefore, LFM is less stable than HSM. The running time of LFM is approximately 1/9 that of HSM, therefore, its calculation efficiency is much higher than that of HSM.