Adaptive sparse polynomial dimensional decomposition for derivative-based sensitivity

Kunkun Tang, Jonathan M. Wang, Jonathan Freund

Research output: Contribution to journalArticle

Abstract

For applications, such as the plasma-coupled combustion system we consider, derivative-based sensitivity indices (DSI) are known to have several advantages over Sobol's total sensitivity indices, especially for small sample sizes. Several properties of derivative-based sensitivity measures are leveraged to develop a new and efficient numerical approach to estimate them. It is based on computing the DSI measures by effectively cost-free Monte Carlo sampling of an adaptively constructed orthogonal polynomial surrogate with uncertain input parameters that can have arbitrary probability distributions. The adaptivity reduces the number of necessary model evaluations, which is demonstrated both in a constructed example (the Moon function) and in two plasma-combustion systems with up to 55 uncertain parameters. Unimportant parameters are successfully identified and neglected with a low number of model evaluations, which makes it an attractive non-intrusive approach when adjoint solutions are unavailable to provide sensitivity information.

Original languageEnglish (US)
Pages (from-to)303-321
Number of pages19
JournalJournal of Computational Physics
Volume391
DOIs
StatePublished - Aug 15 2019

Fingerprint

polynomials
Polynomials
Derivatives
Decomposition
decomposition
sensitivity
Plasmas
Moon
Probability distributions
evaluation
Sampling
moon
sampling
Costs
costs
estimates

Keywords

  • ANOVA
  • Derivative-based sensitivity indices
  • Global sensitivity analysis
  • High dimensionality
  • Non-intrusive sensitivity
  • Polynomial dimensional decomposition (PDD)

ASJC Scopus subject areas

  • Numerical Analysis
  • Modeling and Simulation
  • Physics and Astronomy (miscellaneous)
  • Physics and Astronomy(all)
  • Computer Science Applications
  • Computational Mathematics
  • Applied Mathematics

Cite this

Adaptive sparse polynomial dimensional decomposition for derivative-based sensitivity. / Tang, Kunkun; Wang, Jonathan M.; Freund, Jonathan.

In: Journal of Computational Physics, Vol. 391, 15.08.2019, p. 303-321.

Research output: Contribution to journalArticle

@article{586db866075d4518804eeebf9d391b21,
title = "Adaptive sparse polynomial dimensional decomposition for derivative-based sensitivity",
abstract = "For applications, such as the plasma-coupled combustion system we consider, derivative-based sensitivity indices (DSI) are known to have several advantages over Sobol's total sensitivity indices, especially for small sample sizes. Several properties of derivative-based sensitivity measures are leveraged to develop a new and efficient numerical approach to estimate them. It is based on computing the DSI measures by effectively cost-free Monte Carlo sampling of an adaptively constructed orthogonal polynomial surrogate with uncertain input parameters that can have arbitrary probability distributions. The adaptivity reduces the number of necessary model evaluations, which is demonstrated both in a constructed example (the Moon function) and in two plasma-combustion systems with up to 55 uncertain parameters. Unimportant parameters are successfully identified and neglected with a low number of model evaluations, which makes it an attractive non-intrusive approach when adjoint solutions are unavailable to provide sensitivity information.",
keywords = "ANOVA, Derivative-based sensitivity indices, Global sensitivity analysis, High dimensionality, Non-intrusive sensitivity, Polynomial dimensional decomposition (PDD)",
author = "Kunkun Tang and Wang, {Jonathan M.} and Jonathan Freund",
year = "2019",
month = "8",
day = "15",
doi = "10.1016/j.jcp.2019.04.042",
language = "English (US)",
volume = "391",
pages = "303--321",
journal = "Journal of Computational Physics",
issn = "0021-9991",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Adaptive sparse polynomial dimensional decomposition for derivative-based sensitivity

AU - Tang, Kunkun

AU - Wang, Jonathan M.

AU - Freund, Jonathan

PY - 2019/8/15

Y1 - 2019/8/15

N2 - For applications, such as the plasma-coupled combustion system we consider, derivative-based sensitivity indices (DSI) are known to have several advantages over Sobol's total sensitivity indices, especially for small sample sizes. Several properties of derivative-based sensitivity measures are leveraged to develop a new and efficient numerical approach to estimate them. It is based on computing the DSI measures by effectively cost-free Monte Carlo sampling of an adaptively constructed orthogonal polynomial surrogate with uncertain input parameters that can have arbitrary probability distributions. The adaptivity reduces the number of necessary model evaluations, which is demonstrated both in a constructed example (the Moon function) and in two plasma-combustion systems with up to 55 uncertain parameters. Unimportant parameters are successfully identified and neglected with a low number of model evaluations, which makes it an attractive non-intrusive approach when adjoint solutions are unavailable to provide sensitivity information.

AB - For applications, such as the plasma-coupled combustion system we consider, derivative-based sensitivity indices (DSI) are known to have several advantages over Sobol's total sensitivity indices, especially for small sample sizes. Several properties of derivative-based sensitivity measures are leveraged to develop a new and efficient numerical approach to estimate them. It is based on computing the DSI measures by effectively cost-free Monte Carlo sampling of an adaptively constructed orthogonal polynomial surrogate with uncertain input parameters that can have arbitrary probability distributions. The adaptivity reduces the number of necessary model evaluations, which is demonstrated both in a constructed example (the Moon function) and in two plasma-combustion systems with up to 55 uncertain parameters. Unimportant parameters are successfully identified and neglected with a low number of model evaluations, which makes it an attractive non-intrusive approach when adjoint solutions are unavailable to provide sensitivity information.

KW - ANOVA

KW - Derivative-based sensitivity indices

KW - Global sensitivity analysis

KW - High dimensionality

KW - Non-intrusive sensitivity

KW - Polynomial dimensional decomposition (PDD)

UR - http://www.scopus.com/inward/record.url?scp=85064925109&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064925109&partnerID=8YFLogxK

U2 - 10.1016/j.jcp.2019.04.042

DO - 10.1016/j.jcp.2019.04.042

M3 - Article

VL - 391

SP - 303

EP - 321

JO - Journal of Computational Physics

JF - Journal of Computational Physics

SN - 0021-9991

ER -