Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Risk, Reliability and Safety: Innovating Theory and Practice - Proceedings of the 26th European Safety and Reliability Conference, ESREL 2016 |
| Untertitel | Innovating Theory and Practice - Proceedings of the 26th European Safety and Reliability Conference, ESREL 2016 |
| Herausgeber/-innen | Lesley Walls, Matthew Revie, Tim Bedford |
| Seiten | 50 |
| Seitenumfang | 1 |
| Publikationsstatus | Veröffentlicht - 2017 |
Abstract
Complex engineering systems are usually investigated by running expensive computer models. However, high-dimensional input and computational cost often hinder the calibration of such models. History matching is a calibration strategy, which has been successfully applied in many disciplines. It starts with a specific observation of a physical system, then sequentially discards the implausible input set, that is, the range of input values that do not provide a good fit between the model output and the available data. The computational efficiency depends on Bayesian emulation, which provides a statistical approximation to the model output. At each iteration, the non-implausible domain is sampled to run the simulator until it finds the matching input domain. Because the non-implausible domain can be significantly smaller than the original space, this paper proposes to use an engineering reliability method: Subset Simulation. This strategy uses Markov Chain Monte Carlo to sample from a failure domain. Numerical examples are provided to show the potential of this sampling scheme for model calibration.
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Risk, Reliability and Safety: Innovating Theory and Practice - Proceedings of the 26th European Safety and Reliability Conference, ESREL 2016: Innovating Theory and Practice - Proceedings of the 26th European Safety and Reliability Conference, ESREL 2016. Hrsg. / Lesley Walls; Matthew Revie; Tim Bedford. 2017. S. 50.
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - Bayesian model calibration using subset simulation
AU - Gong, Z.T.
AU - DiazDelaO, F.A.
AU - Beer, M.
N1 - Publisher Copyright: © 2017 Taylor & Francis Group, London.
PY - 2017
Y1 - 2017
N2 - Complex engineering systems are usually investigated by running expensive computer models. However, high-dimensional input and computational cost often hinder the calibration of such models. History matching is a calibration strategy, which has been successfully applied in many disciplines. It starts with a specific observation of a physical system, then sequentially discards the implausible input set, that is, the range of input values that do not provide a good fit between the model output and the available data. The computational efficiency depends on Bayesian emulation, which provides a statistical approximation to the model output. At each iteration, the non-implausible domain is sampled to run the simulator until it finds the matching input domain. Because the non-implausible domain can be significantly smaller than the original space, this paper proposes to use an engineering reliability method: Subset Simulation. This strategy uses Markov Chain Monte Carlo to sample from a failure domain. Numerical examples are provided to show the potential of this sampling scheme for model calibration.
AB - Complex engineering systems are usually investigated by running expensive computer models. However, high-dimensional input and computational cost often hinder the calibration of such models. History matching is a calibration strategy, which has been successfully applied in many disciplines. It starts with a specific observation of a physical system, then sequentially discards the implausible input set, that is, the range of input values that do not provide a good fit between the model output and the available data. The computational efficiency depends on Bayesian emulation, which provides a statistical approximation to the model output. At each iteration, the non-implausible domain is sampled to run the simulator until it finds the matching input domain. Because the non-implausible domain can be significantly smaller than the original space, this paper proposes to use an engineering reliability method: Subset Simulation. This strategy uses Markov Chain Monte Carlo to sample from a failure domain. Numerical examples are provided to show the potential of this sampling scheme for model calibration.
UR - http://www.scopus.com/inward/record.url?scp=85016194023&partnerID=8YFLogxK
M3 - Conference contribution
SN - 9781138029972
SP - 50
BT - Risk, Reliability and Safety: Innovating Theory and Practice - Proceedings of the 26th European Safety and Reliability Conference, ESREL 2016
A2 - Walls, Lesley
A2 - Revie, Matthew
A2 - Bedford, Tim
ER -