On the identification of damping from non-stationary free decay signals using modern signal processing techniques

A numerical implementation of system identification from non-linear and non-stationary signals is presented. The continuous wavelet transform (CWT) along with the complex Morlet wavelet skeleton curve extraction and Hilbert Transform (HT)-based methodologies are used for identification purposes. A comparison of the advantages of each technique in the analysis of non-stationary free decay systems is presented and improvements to the current methodologies are proposed. The HT approach offered good results in the estimation of the instantaneous amplitude in low damping and non-noisy signals. However, it is highly sensitive to impulses and irregularities in the signal, which affects the proper detection of frequency and amplitude parameters in real-life signals. The CWT exhibited better results for the analysis of noisy signals, from the resulting wavelet map the noise content can be distinguished from the actual system response. That is, the modes show a distinctive pattern in the map allowing proper modal extraction. However, for highly damped non-stationary decaying signals, the results are affected by the decay rate, round-up errors, and edge effects.