In this work, a signal processing method based on the Empirical Mode Decomposition (EMD) to denoise a recorded signal is proposed. EMD expresses the signal as an expansion of basis functions (Intrinsic Mode Functions - IMFs) that are signal dependent and are estimated via an iterative procedure.The decomposition of an "only noise" signal is first studied to define a Noise-Model in terms of energy and period. Then, the EMD is applied to a simulated measured signal, and the IMFs obtained are compared with the Noise-Model constructed before. Finally, an optimization procedure is performed to split the IMFs of the measured signal into 2 components: The denoised IMFs and the corresponding "Removed Noise" IMFs. The denoised IMFs are finally summed in order to reconstruct the denoised signal. The proposed algorithm is applied to a simple 3-floor shear-type frame and the ASCE 4-floor frame benchmark. The results are compared with those obtained by a standard denoising procedure based on a pass-band filter; the comparison confirmed the improvements obtained with the proposed method over classical procedures. © 2013 Taylor & Francis Group, London.
Mukhopadhyay S., Betti R., Galli E., Savoia M., Vincenzi L. (2013). A new denoising procedure based on empirical mode decomposition for SHM purpose. George Deodatis, Bruce R. Ellingwood, Dan M. Frangopol.
A new denoising procedure based on empirical mode decomposition for SHM purpose
BETTI, RICCARDO;GALLI, ELISA;SAVOIA, MARCO;VINCENZI, LORIS
2013
Abstract
In this work, a signal processing method based on the Empirical Mode Decomposition (EMD) to denoise a recorded signal is proposed. EMD expresses the signal as an expansion of basis functions (Intrinsic Mode Functions - IMFs) that are signal dependent and are estimated via an iterative procedure.The decomposition of an "only noise" signal is first studied to define a Noise-Model in terms of energy and period. Then, the EMD is applied to a simulated measured signal, and the IMFs obtained are compared with the Noise-Model constructed before. Finally, an optimization procedure is performed to split the IMFs of the measured signal into 2 components: The denoised IMFs and the corresponding "Removed Noise" IMFs. The denoised IMFs are finally summed in order to reconstruct the denoised signal. The proposed algorithm is applied to a simple 3-floor shear-type frame and the ASCE 4-floor frame benchmark. The results are compared with those obtained by a standard denoising procedure based on a pass-band filter; the comparison confirmed the improvements obtained with the proposed method over classical procedures. © 2013 Taylor & Francis Group, London.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
