The paper presents a comparative study of the effectiveness of three novel damage detection techniques namely Continuous Wavelet Transform (CWT), Empirical Mode Decomposition (EMD) and Wavelet Packet Sifting (WPS). The health condition of a mechanical or civil engineering structure can be assessed by monitoring a change in natural frequencies and mode shapes. CWT method can be used to identify the instantaneous values of these modal parameters by the wavelet ridges. Using the EMD method, intrinsic mode functions (IMF) can be sifted from a vibration signal, whereas a newly-developed WPS technique can decompose a signal into its dominant mono-frequency components. Instantaneous modal information can be extracted by incorporating the EMD and WPS with the Hilbert Transform. These techniques are illustrated for simulated vibration data from a three-degree-of-freedom system subjected to (i) sudden damage and (ii) progressive damage. The aspects related to the implementation algorithms, sensitivity to damage type and the robustness issues in case of noisy data are discussed. In case of progressive damage, all methods performed well. WPS technique performed better in case of sudden damage whereas CWT demonstrated robustness in case of noisy data.
On transfer learning for chatter detection in turning using wavelet packet transform and ensemble empirical mode decomposition
