Probabilistic damage identification using improved approximate Bayesian computation

摘要：

　　Uncertainties always exist in real-world engineering structures resulting in inapplicability of deterministic damage identification methods.To overcome the limitation of classic Bayesian inference in probabilistic damage identification,this paper proposes an improved approximate Bayesian method incorporating approximate Bayesian computation,the Metropolis Hastings sampling and stochastic response surface.The approximate Bayesian computation avoids solving the likelihood function in the Bayesian formula,which considerably simplifies the posterior probability estimation of uncertain parameters.The Metropolis Hastings sampling provides convergent and stable samples for accurate estimation of parameter posterior probabilities.Meanwhile,stochastic response surface establishes the explicit mathematical expression between uncertain parameters and responses,which highly improves the efficiency in estimating statistical features of responses.Finally,the feasibility of the proposed method has been validated using a numerical reinforced concrete beam whose damage was identified by comparing the parameter posterior distributions before and after damage.

关键词： Probabilistic damage identification approximate Bayesian computation Metropolis Hastings sampling stochastic response surface probability distributions

作者: Sheng-En FANG Shan CHEN

作者单位: School of Civil Engineering,Fuzhou University,Fuzhou,China;National and Local United Research Center School of Civil Engineering,Fuzhou University,Fuzhou,China

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: 1970-1976

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）