A Hierarchical Bayesian Framework for System Identification of Dynamical Systems

摘要：

　　From a Bayesian perspective,the probability is defined as the relative plausibility of a particular outcome conditional on an observed set of data such that the probability axioms are appeared satisfied.This extended concept of probability allows it updating models and their uncertainties according to a limited amount of information.In practice,as models are often misspecified,the Bayesian estimations suffer from significant variability when they are inferred from independent experiments.The present Bayesian methods do not account for such variabilities.Consequently,the outcome of Bayesian methods is often criticized for being inconsistent with the second moment statistics computed using a frequentist approach.In this paper,this problem is critically reviewed and addressed using a new hierarchical Bayesian framework.The proposed framework unifies the two probabilistic paradigms aiming at enhancing the accuracy of uncertainty quantification.It is proved that the uncertainty computed by the hierarchical Bayesian approach is consistent with the frequentist approach,although the uncertainty is justified on the grounds of the Bayesian probability logic.

关键词： Bayesian approach Frequentist approach Model updating Uncertainty Quantification Hierarchical models

作者: O.Sedehi L.S.Katafygiotis C.Papadimitriou

作者单位: Department of Civil and Environmental Engineering,The Hong Kong University of Science and Technology Department of Civil and Environmental Engineering,The Hong Kong University of Science and Technology Department of Mechanical Engineering,University of Thessaly,Volos,Greece

会议类型: 国际会议

会议名称: The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)

会议地点: 青岛

会议语种:英文

页码: 1812-1819

在线出版日期: 2018-07-22（万方平台首次上网日期，不代表论文的发表时间）