A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
An Efficient Markov Chain Monte Carlo Method for Bayesian System Identification of Tower Structures
Abstract Treating system identification as a Bayesian inference problem, this paper proposes an efficient Markov chain Monte Carlo method for identifying the posterior PDF of uncertain system parameters, given measured data. Instead of pinpointing a model, the proposed method considers all the models in the parameter space with the posterior PDF describing their relative plausibility. The Markov chain Monte Carlo method is used to sample from the posterior PDF. In particular, a multi-level sampling scheme based on sequential importance sampling is applied to fully explore the high-dimension parameter space, so that all the multiple isolated peaks of the posterior PDF for the identifiable case, and the extended and highly complex manifold of the posterior PDF for the unidentifiable case can be captured. By investigating the mathematical structure of the problem, a proposal PDF of the Metropolis-Hastings algorithm is developed for efficiently sampling from the posterior PDF. The proposed method is illustrated with simulated data. Apart from system identification, the proposed method also finds its applications in other Bayesian inference problems.
An Efficient Markov Chain Monte Carlo Method for Bayesian System Identification of Tower Structures
Abstract Treating system identification as a Bayesian inference problem, this paper proposes an efficient Markov chain Monte Carlo method for identifying the posterior PDF of uncertain system parameters, given measured data. Instead of pinpointing a model, the proposed method considers all the models in the parameter space with the posterior PDF describing their relative plausibility. The Markov chain Monte Carlo method is used to sample from the posterior PDF. In particular, a multi-level sampling scheme based on sequential importance sampling is applied to fully explore the high-dimension parameter space, so that all the multiple isolated peaks of the posterior PDF for the identifiable case, and the extended and highly complex manifold of the posterior PDF for the unidentifiable case can be captured. By investigating the mathematical structure of the problem, a proposal PDF of the Metropolis-Hastings algorithm is developed for efficiently sampling from the posterior PDF. The proposed method is illustrated with simulated data. Apart from system identification, the proposed method also finds its applications in other Bayesian inference problems.
An Efficient Markov Chain Monte Carlo Method for Bayesian System Identification of Tower Structures
Yang, Jia-Hua (author) / Lam, Heung-Fai (author)
2019-09-04
9 pages
Article/Chapter (Book)
Electronic Resource
English
Bayesian Inference and Markov Chain Monte Carlo Methods
| Wiley | 2013
Slope Stability Analysis Using Bayesian Markov Chain Monte Carlo Method
| Online Contents | 2020
Decision making of HVAC system using Bayesian Markov chain Monte Carlo method
| Online Contents | 2014