3rd-order Spectral Representation Method: Part II -- Ergodic Multi-variate random processes with fast Fourier transform

Abstract

The second in a two-part series, this paper extends the 3rd-order Spectral Representation Method for simulation of ergodic multi-variate stochastic processes according to a prescribed cross power spectral density and cross bispectral density. The 2nd and 3rd order ensemble properties of the simulated stochastic vector processes are shown to satisfy the target cross correlation properties in expectation. A multi-indexed frequency discretization is introduced to ensure ergodicity of the sample functions. This is first shown for uni-variate processes and then the simulation formula for multi-variate processes is provided. Ensemble properties and ergodicity of the sample functions are proven. Additionally, it is shown that the simulations can be implemented efficiently with the Fast Fourier Transform, which greatly reduces computational effort. An example involving the simulation of turbulent wind velocity fluctuations is presented to further highlight the features and applications of the algorithm.