DENOISING CORRUPTED STRUCTURAL VIBRATION RESPONSE: CRITICAL COMPARISON AND ASSESSMENT OF RELATED METHODS

Abstract

Vibration-based Structural Health Monitoring (SHM) is receiving increasing attention due to the high technological level that the developed methodologies are nowadays reaching, together with the increasing quest for the implementation of effective, but ideally of low-cost, monitoring instrumentation. The latter is relatively easy to deploy and allows for the recording of the structural vibration response at multiple locations, which, for large strategical infrastructures, such as high-rise buildings, bridges, wind farms, etc. may be of critical importance. On the other hand, low-cost instrumentation may typically be accompanied by high Noise-to-Signal (N/S) ratios, contaminating the structural response, increasing the induced uncertainties and rendering the implementation of SHM methods more difficult. This calls for appropriate and effective denoising processes. In tackling such an associated denoising problem, several methods have been proposed and are currently under further development. Among specific variants, the utilization of multi-rate filter banks, especially the one based on Discrete Wavelet Transform (DWT), as well as the application of Singular Value Decomposition (SVD), have revealed to be rather effective. Yet, they have mostly been applied to problems where structural vibration response signals originate from specific behavioral classes, as, e.g., in monitoring applications of rotating machinery. In this study, the aforementioned methods are reconsidered and reimplemented; then, assessed on noise-corrupted vibration response signals related to civil engineering applications. Different Noise-to-Signal (N/S) ratios and excitation types are considered, i.e. earthquake and ambient vibration input. Advantages and limitations of both denoising approaches are presented and discussed, with a critical comparison. The outcomes prove the effectiveness of the considered methods in clarifying earthquake response signals, yet also their difficulty in purifying ambient vibration signals, motivating further specific research on the filtering of this latter common class of structural response.