BAYESIAN-BASED NONLINEAR MODEL UPDATING AND DYNAMIC RELIABILITY ANALYSIS

A Bayesian-based method for nonlinear model updating is proposed. Considering the randomness of excitation, a dynamic reliability analysis method of compound random vibration systems is established. The instantaneous characteristic parameters of the principal components decomposed from the measured structural dynamic response are considered as the nonlinear index that is applied to construct the likelihood function. The DRAM-Gaussian process model combined method is further performed for the uncertainty quantification of the calibrated model parameters. According to the first transcendental failure criterion, the generalized probability density evolution method is used to calculate the structural dynamic reliability. Both the deterministic nominal model and the compound random vibration system are considered. The accuracy of the calculated reliability is subsequently verified by Monte Carlo simulation. The results show that the Bayesian-based measured instantaneous characteristics can be used to quantify the uncertainty of the nonlinear model parameters with the advantages of high accuracy and efficiency. In addition, the dynamic reliability of the compound random model that considers the uncertainty of structural parameters is generally lower than that of the deterministic nominal model. Thus, it will make the evaluation results safer and more reliable by considering the effects of model parameters uncertainties in structural safety state assessment.