Online detection of time-variant oscillations based on improved ITD

An online detector for time-variant oscillations in univariate time-series is proposed. This paper is motivated by the fact that it is still an open problem to design a real-time oscillation detector which is applicable to non-linear, non-stationary and intermittent oscillations. The proposed procedure is based on intrinsic time-scale decomposition (ITD) and contains (i) an improved iteration termination condition for ITD with on-line back-redecomposition and (ii) a novel hypothesis test with a robust statistic of variation coefficient which enables online monitoring of time-variant oscillations. The proposed approach is computationally efficient, does not require a priori supervision window and is better applicable for online detection of time-variant oscillations. In addition, it preserves nonlinear features in process variables which facilitates subsequent oscillation diagnosis. Simulation examples and industrial applications are provided to demonstrate the effectiveness of the proposed online oscillation detector.