Reliability updating of non-linear dynamic system using vibration data

Abstract

Reliability updating has always been an important field of study in structural health monitoring due to the existence of structural properties’ errors. These errors may lead to incorrect prediction of structural behaviors. Reliability updating allows the accurate structural properties to be obtained by statistical method. The statistical method is based on the framework of Bayesian Statistics which is able to update the possible value of the structural properties through the prior belief and its likelihood (evidence). However, in reality the building structure always have high dimensional parameters. Using Bayes’ theorem to solve for such problem can lead to an integration of multivariate which will requires significantly computational efforts. Therefore, Subset Simulation is brought in to solve high dimensional model updating problems. Moreover, the computation of the failure probability of the structures requires enormous number of samples to be generated since the failure probability can be extremely small. Utilization of Subset Simulation breaks the failure event down to a series of intermediate events with a relatively high probability which can greatly reduce the time in sample generation.