Computation of the Sobol' Indices using importance sampling

Pierre Beaurepaire, Matteo Broggi, Edoardo Patelli

Research output: Chapter in Book/Report/Conference proceedingConference contribution book

Abstract

The contribution of uncertain parameters to the variability of a model may be estimated by means of global sensitivity analysis. The use of the first order Sobol' indices is relevant in this context. However, accurate estimation of these coefficients may be a challenging problem, since as many Monte Carlo simulations as the dimension of the input parameters of the problem have to be performed in order to compute all the first order Sobol' indices. Such numerical efforts are not practicable for many large scale applications. This paper presents a novel procedure which requires to perform a single Monte Carlo simulation to determine all the Sobol' indices associated with the input parameters. The benefits of the method are illustrated with an application example.

Original languageEnglish
Title of host publicationVulnerability, Uncertainty, and Risk
Subtitle of host publicationQuantification, Mitigation, and Management - Proceedings of the 2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014
EditorsJim W. Hall, Siu-Kui Au, Michael Beer
PublisherAmerican Society of Civil Engineers (ASCE)
Pages2115-2124
Number of pages10
ISBN (Electronic)9780784413609
DOIs
Publication statusPublished - 7 Jul 2014
Event2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014 - Liverpool, United Kingdom
Duration: 13 Jul 201416 Jul 2014

Publication series

NameVulnerability, Uncertainty, and Risk: Quantification, Mitigation, and Management - Proceedings of the 2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014

Conference

Conference2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014
CountryUnited Kingdom
CityLiverpool
Period13/07/1416/07/14

Keywords

  • computing in civil engineering
  • sensitivity analysis
  • variance-based sensitivity
  • uncertainty analysis
  • importance sampling
  • intelligent systems
  • Monte Carlo methods
  • parameter estimation
  • risk analysis
  • risk assessment

Fingerprint Dive into the research topics of 'Computation of the Sobol' Indices using importance sampling'. Together they form a unique fingerprint.

  • Cite this

    Beaurepaire, P., Broggi, M., & Patelli, E. (2014). Computation of the Sobol' Indices using importance sampling. In J. W. Hall, S-K. Au, & M. Beer (Eds.), Vulnerability, Uncertainty, and Risk: Quantification, Mitigation, and Management - Proceedings of the 2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014 (pp. 2115-2124). (Vulnerability, Uncertainty, and Risk: Quantification, Mitigation, and Management - Proceedings of the 2nd International Conference on Vulnerability and Risk Analysis and Management, ICVRAM 2014 and the 6th International Symposium on Uncertainty Modeling and Analysis, ISUMA 2014). American Society of Civil Engineers (ASCE). https://doi.org/10.1061/9780784413609.212