Digital Generation of Non-Gaussian Spiky Excitations Using Spectral Representation with Additive Phase Structure

Abstract

AbstractThis paper presents a framework of the digital generation of non-Gaussian spiky excitations. This study is focused on the random spikiness, featuring large excursions with considerable energy and monotonic (nonstochastic) variations in a local time history. A first-order non-Gaussian stochastic time series model and its spectral representation are employed for the local spiky features. The stochastic model generates not only autocorrelation properties but also a unique shape of peaks formed with random spikes and monotonic variations between spikes. The Fourier representation of the stochastic model enables an effective control of the peaks and provides a filtering operation for the local feature generation in the frame of stationary stochastic process. Several spectral models with stochastic or ensemble-averaged amplitudes and four added phase functions have been developed. Thus, the phase is different from the uncorrelated uniform phases in a conventional spectral method. The essential feature o...