MODAL PARAMETER IDENTIFICATION OF AN ACTUAL BRIDGE BY FAST BAYESIAN FFT METHOD UNDER AMBIENT EXCITATION

The Bayesian theory is adopted in modal parameter identification, finite element model updating and condition assessment of the structures recently. Bayesian modal identification can be preformed quickly using Fast Bayesian FFT method where most probable values (MPVs) are obtained by solving at most a four-dimensional numerical optimization and Hessian matrix is calculated by second derivatives of log-likelihood function with respect to modal parameters for a single modal in a selected resonant frequency band. In this paper, in order to evaluate its feasibility and superiority for modal parameter identification of actual bridges, Fast Bayesian FFT method is adopted to process the vertical acceleration response of a rigid frame-continuous girders bridge under ambient excitation. In addition to the best estimates of modal parameters, their posterior uncertainties were also assessed by their coefficients of variation. Comparison with the results from stochastic subspace identification (SSI) indicates that identified modal frequencies and mode shapes agree very well, but the variation of the identified modal damping ratios is still high.