A frequency domain approach for wide-band non-Gaussian process

A number of frequency domain approaches for wide-band Gaussian process have been proposed and some of them have succeeded in estimating fatigue damage with high accuracy. However, in many real engineering problems, it is not hard to find structural responses exhibiting non-Gaussian properties. As non-Gaussianity has an effect on statistical properties and fatigue damage, the existing techniques based on the Gaussian assumption should be modified to reflect the non-Gaussianity. Among them, Tovo-Benascuitti model provided a framework to extend probability density function derived in Gaussian assumption to non-Gaussian area. This model gave a good result in describing probability density of stress amplitude but estimated probability density function of stress mean seemed to be deviated from real data. The errors of this model in probability density of stress mean increased the error of estimated fatigue damages. In this work, conditional density of stress mean given stress amplitude is assumed to be zero-mean Gaussian function for reflecting realistic information of stress mean in joint probability density function of stress mean and amplitude. The estimated fatigue damages calculated by the modified model in non-Gaussian process are significantly improved. The performance of proposed model is addressed in numerical simulations.