Spectral method for non-Gaussian data generation by phase modeling: White noise phase versus structured phase

• New spectral framework based on non-white-noise phases.
• Spectral representation of stochastic peak generation model.
• Non-Gaussian property simulation by phase modeling.
• Simulation capability of high-resolutuion spiky time histories.

This paper presents a spectral method for non-Gaussian data generation by phase modeling. Independent random phases have been used in a standard spectral method. Thus, the use of structured phases for non-Gaussian series in EARPG(1)/UPS is innovative but appears to conflict with the conventional modeling concept. This paper revisits the method and clarifies a difference in framework, which is apparently the white noise phase versus the structured phase but essentially a change in the whole framework. This paper rebuts the claim by Kumar and Stathopoulos in this journal (Suresh Kumar and Stathopoulos, 1999) that EARPG(1)/UPS using structured phases is unnecessarily complicated and inefficient compared with their EPG considering phase whitening, which is misleading in regard to the new modeling concept and the major simulation capability. The rebuttal clarifies that the new approach is based on a non-conventional time series modeling focused on spiky peak generation and a filtering ability of Fourier representation of the model. This new concept establishes a new spectral framework suitable for large-scale spiky time history simulation, permitting local non-stochastic variations and high-resolution sharpness in addition to the statistical properties of usual time series models. Simulation of wind-induced excursion loading is demonstrated.