Analysis of Process Dynamics with Monte Carlo Singular Spectrum Analysis

Singular spectrum analysis of time series observations can be visualized as a sliding window of width m moving down the time series x of length n, to determine the orthogonal patterns that best capture the variance presented by the window. It is designed to extract information from short and noisy time series. This allows the time series to be decomposed into various components, ideally separating components containing information from those that contain mostly noise. In this paper, Monte Carlo singular spectrum analysis is used to separate the components of time series data obtained from several real world processes into signal and noise components. The methodology facilitates better understanding of the underlying dynamics of complex systems, as well as the development of automated plant systems.