Wavelet-based vibration denoising for structural health monitoring

In the context of civil engineering applications, vibration responses are complex, exhibiting variations in time and space and often containing nonlinearity and uncertainties not considered during data collection. These responses can also be contaminated by various sources, impacting damage identification processes. A significant challenge is how to effectively remove noise from these data to obtain reliable damage indicators that are unresponsive to noise and environmental factors.This study proposes a new denoising algorithm based on discrete wavelet transform (DWT) that addresses this issue. The suggested method offers a strategy for denoising using distinct thresholds for positive and negative coefficient values at each band and applying denoising process to both detail and trend components. The results prove the effectiveness of the technique and show that Bayes thresholding performs better than the other techniques in terms of the evaluated metrics. This suggests that Bayes thresholding is a more accurate and robust technique for thresholding compared to other common techniques.