On frequency detection and root cause analysis of plant-wide oscillations via ACF method

This paper addresses the problem of detecting plant wide oscillations in industrial processes and additionally to realize the root cause analysis. When raw signals with the presence of noise, non periodic perturbations, and very low frequencies are analyzed, even robust algorithms may fail to detect oscillating signals. In this paper, a well know algorithm based on measuring the period of filtered covariance from power spectrum density plots is analyzed and improvements are made to handle the above situations, so that raw signals can be correctly evaluated without the need of customized parameterizations. The measurement of variability of filtered signals is also proposed for root cause analysis. The contributions of this paper are highlighted through its application to data of four different processes.