Numerical Simulation Of Homogeneous Non-Gaussian Random Vector Fields

Abstract

A method which constructs numerical simulations of homogeneous non-Gaussian random vector fields is given. It uses multi-dimensional pulse trains with Poisson arrival times. It is first shown that the classical results concerning one-dimensional filtered Poisson processes can be generalized to the multi-dimensional case. It is then explained how to determine the multi-dimensional filtered Poisson process in order to match the spectral density matrix and the first statistical moments of the given non-Gaussian random field. This method is used in order to construct non-Gaussian simulation of wind. Finally, the sensitivity of the response of a non-linear dynamical system excited by a random process to the input law is illustrated by an example.