Vector outcrossing probabilities by Monte Carlo

Abstract An exact method for calculating the probability that a vector-valued Gaussian stationary process will cross out of a safe set in a given time, is presented. It is based on representing the process by a random trigonometric polynomial and then calculating the probability by directional simulation. Exact bounds for the error can be calculated, giving a predetermined accuracy. A numerical example relating to the response of a structure to wind excitation is given.