Bayesian wavelet probabilistic method for nonparametric damage detection of structures

This paper presents a Bayesian wavelet probabilistic methodology for nonparametric damage detection of building structures under external excitations. A Bayesian discrete wavelet packet transform-based denoising approach is applied for data cleansing. A nonparametric fuzzy wavelet neural networks model is employed to predict structural dynamical response. Structural condition is evaluated using Bayesian hypothesis testing approach. The Bayesian metric is treated as a random variable and its probability density function is simulated using Monte Carlo technique. The methodology is validated using the sensed data collected from a 5-story test steel frame. Both the original and denoised data are used in the damage detection to investigate the effects of noise on detection accuracy.