Structural response reconstruction based on the information fusion of multi-source particle filters

Aiming at the problems that the lack of theoretical basis for the selection of particle set sampling variance and the resampling methods in traditional particle filter algorithms, and sampling process is easily disturbed by noise, an uncertainty structural response reconstruction method based on the information fusion of multi-source particle filters is proposed. Firstly, the sampling variance of particle set is analogous to the accuracy index of sensors, and a number of independent particle filtering samples from different sources are performed to ensure the independence of particles. Then, abnormal filters are screened and eliminated according to relative percentage error (RPE) threshold of preliminary reconstruction, and the state estimation results of remained particle filters are fused by the multi-source sensors information fusion technique to approximate to the real state values with high accuracy. Finally, the fused state values and the state space models are employed to reconstruct the responses of key positions, and the effectiveness of the proposed method is verified by numerical example of the space truss structure and the cantilever beam test. The results show that the proposed method can reduce the influence of the above uncertainties on reconstruction results, effectively improve the particle impoverishment problem, the filtering stability is good and the reconstruction accuracy is high.