A New Denoising Technique via Wavelet Analysis of Structural Vibration Response for Structural Health Monitoring Applications

In various civil engineering applications (CEAs), most vibration responses vary with time and space and are characterized by nonlinearities and uncertainties that are not accounted for during data acquisition. Furthermore, these responses may be contaminated by various sources, which may affect the damage identification process. The main challenge is how to denoise these data in order to acquire a sensitive feature for damage identification that is insensitive to noise and environmental effects. Wavelet Transform (WT) has been proven to be useful for denoising in the field of structural health monitoring (SHM). However, its efficiency is affected by the selection of wavelet parameters. The questions related to the best approach for utilizing the most suitable parameters have not been adequately answered. This study attempts to address this issue by proposing a new denoising algorithm based on the Discrete Wavelet Transform (DWT) technique. The proposed technique provides a strategy to choose the right decomposition levels for denoising and selecting proper mother wavelets. The proposed algorithm uses separate noise thresholds for negative and positive coefficients at each level and applies denoising to detail and approximation components. Datasets from actual civil structures have been analyzed. According to the presented experimental results, the proposed technique exhibits promising results for signal denoising using "db3" compared with traditional techniques. In addition, "db3" and "sym3" are shown to be the best choices for the mother wavelet.