Signal denoising method for modal analysis of an offshore platform

The modal analysis of offshore structures is a key element of structural health monitoring based on vibration. A difficulty encountered by practitioners and researchers is that an accurate modal analysis is challenging in noisy environments. The objective of this work is to develop and implement a signal denoising method based on solving the inverse singular value problem of a measured (noisy) data matrix with prescribed entries to reconstruct a filtered data matrix. The measured (noisy) impulse response function (IRF) is used to build a square or nearly square Hankel matrix. The normalized rank determination indicator (RDI) of the Hankel matrix is introduced to determine the number of prescribed non-zero singular values in the prescribed entries inverse singular value problem (PEISVP). The reconstructed (filtered) matrix must maintain the original Hankel structure, and preserve the number of non-zero singular values. Once the filtered IRF has been obtained, the complex exponential (CE) method is applied for the modal analysis. To validate the proposed method, hereafter referred to as PEISVP-CE, we undertake the numerical simulation of a five-story shear building. Once successfully validated, we apply the PEISVP-CE method to an offshore field experiment of a jacket platform under the step relaxation. We find that the PEISVP-CE method is effective at eliminating noise and, therefore, appropriate for the modal analysis of offshore structures.