ABSTRACT
In this study, a novel parameter identification method based on data fusion algorithm is proposed. A set of sparse matrix equations on state variables is obtained by establishing the state space system of bridge structure parameters. Data passed into the equation is multi-source and heterogeneous, which includes time-history measurements of acceleration and velocity, and displacement estimation acquired from digital image correlation. Bayesian inference with multi-process modification is applied to solve the equation. With the capability of eliminating ill-conditioned equation, method is verified through shake table test. Applying the method to a series of raw test data, results show that estimation error of the identified seismic excitation time history is at a very low level, which represents good effectiveness. Material parameters calculated from the excitation recognition results of the tested specimen are also in good agreement with estimated results, indicating feasibility in practice.
