Detection method of multiple oscillations based on wavelet analysis

An automated detection method is proposed for the detection and assessment of multiple oscillations in process, which can be applied to semi-stationary process as well. Data pre-processing procedure and iterative optimization contribute to the automation of the detection method. Wavelet analysis reveals a better power distribution map than Fourier transform. The map is utilized to reveal the power distribution in time-frequency domain and to design reliable band-pass filters. Filtered auto-covariance function is calculated to analyze the period and regularity of each oscillating component. An iterative optimization algorithm to this method is discussed and industrial data sets are tested to verify the method.