Evaluation of Automated Steady State and Transient State Detection Algorithms
Abstract
A comprehensive comparison of two statistical methods for automated identification of probable steady state and probable transient state in a noisy process signal is performed. Both approaches use the R-statistic method, which is the ratio of estimated variances, for steady state identification and are independent of noise variance. The performance of both approaches is determined based on probability of occurrence of Type-I, Type-II errors and the Average Run Length (ARL) at points of change in the process signal. The effectiveness of both approaches with respect to computational burden, computational time, ease of understanding, storage, etc. is analyzed for step change as well as ramp change in the noisy process signal.
Collections
- OSU Theses [15752]