Online model parameter updating is an effective method to improve the model accuracy of numerical substructure in the hybrid simulation. In order to improve the precision of online parameters identification for the nonlinear model, an improved auxiliary unscented particle filter (AUPF) was proposed in this paper base on the standard particle filter algorithm. In the stage of the importance sampling, the unscented Kalman filter (UKF) method was adopted to calculate estimates of all particles using the latest observation information in order to improve estimation precision of the nonlinear transformation of the particles. In the stage of the resampling, an auxiliary factor was introduced to modify particle weights, which enriched the particle diversity and weaken the particle degradation phenomenon. Parameters online identification for the Bouc-Wen were conducted with the AUPF, and both the identification precision and the calculation efficiency of results were compared with the standard particle filter algorithm (PF), The rextended Kalman particle filter algorithm (EPF), and unscented particle filter algorithm (UPF). The results show that compared with the other three algorithms, the proposed AUPF algorithm can effectively improve the precision of online parameters identification of the Bouc-Wen model and reduce the fluctuation of parameters identification values on the basis of increasing computing time-consuming in a single step. Finally, the effectiveness of the Bouc-Wen model parameter identification method using the AUPF algorithm was verified through the quasi-static test of the rubber isolation bearing.