Damage Detection Based on Bayes Factor Using Ambient Vibration Data

Abstract

This paper presents the work on developing a frequency domain method for damage detection using ambient vibration data. Bayesian factor is used to construct a new damage detection indicator utilizing the concept of Fast Bayesian FFT (Fast Fourier Transform) method. Based on Bayes factor and the properties of FFT data, the prior probability density function (PDF) and the likelihood function in the indicator can be constructed according to Gaussian distribution. The most probable value (MPV) of modal parameters mainly including natural frequency, damping ratio and mode shape, and the corresponding covariance matrix, which can be determined by the Fast Bayesian FFT method, were also used for the development of damage indicator. The uncertainty of modal parameters can be taken into account in the damage detection. A simply supported bridge with 10 elements was simulated to illustrate the proposed method by generating data in different damage cases. The damage location and degree can be identified by instrumenting the sensors in different locations.