Mitigating Error and Uncertainty in Partitioned Analysis: A Review of Verification, Calibration and Validation Methods for Coupled Simulations

Garrison Stevens, Sez Atamturktur

Research output: Contribution to journalArticle

Abstract

Partitioned analysis involves coupling of constituent models that resolve different scales or physics by allowing them to exchange inputs and outputs in an iterative manner. Through partitioning, simulations of complex physical systems are becoming evermore present in the scientific modeling community, making the Verification and Validation (V&V) of partitioned models to quantifying the predictive capability of their simulations increasingly important. Partitioning presents unique challenges, as well as opportunities, for the V&V community. Verification gains a new level of complexity in partitioned models, as numerical errors can easily be introduced at the coupling interface where non-matching domains and models are integrated together. For validation, partitioned analysis allows the quantification of the uncertainties and errors in constituent models through comparison against separate-effect experiments conducted in independent constituent domains. Such experimental validation is important as uncertainties and errors in the predictions of constituents can be transferred across their interfaces, either compensating for each other or accumulating during iterative coupling operations. This paper reviews published literature on methods for assessing and improving the predictive capability of strongly coupled models of physical and engineering systems with an emphasis on advancements made in the last decade.

Original languageEnglish (US)
Pages (from-to)557-571
Number of pages15
JournalArchives of Computational Methods in Engineering
Volume24
Issue number3
DOIs
StatePublished - Jul 1 2017

Fingerprint

Calibration
Uncertainty
Simulation
Partitioning
Model
Verification and Validation
Literature Review
Coupled Model
Experimental Validation
Systems Engineering
Quantification
Resolve
Systems engineering
Physics
Review
Prediction
Output
Modeling
Experiment
Experiments

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Applied Mathematics

Cite this

@article{44653f4a72474192b1cd173cfa81ef4e,
title = "Mitigating Error and Uncertainty in Partitioned Analysis: A Review of Verification, Calibration and Validation Methods for Coupled Simulations",
abstract = "Partitioned analysis involves coupling of constituent models that resolve different scales or physics by allowing them to exchange inputs and outputs in an iterative manner. Through partitioning, simulations of complex physical systems are becoming evermore present in the scientific modeling community, making the Verification and Validation (V&V) of partitioned models to quantifying the predictive capability of their simulations increasingly important. Partitioning presents unique challenges, as well as opportunities, for the V&V community. Verification gains a new level of complexity in partitioned models, as numerical errors can easily be introduced at the coupling interface where non-matching domains and models are integrated together. For validation, partitioned analysis allows the quantification of the uncertainties and errors in constituent models through comparison against separate-effect experiments conducted in independent constituent domains. Such experimental validation is important as uncertainties and errors in the predictions of constituents can be transferred across their interfaces, either compensating for each other or accumulating during iterative coupling operations. This paper reviews published literature on methods for assessing and improving the predictive capability of strongly coupled models of physical and engineering systems with an emphasis on advancements made in the last decade.",
author = "Garrison Stevens and Sez Atamturktur",
year = "2017",
month = "7",
day = "1",
doi = "10.1007/s11831-016-9177-0",
language = "English (US)",
volume = "24",
pages = "557--571",
journal = "Archives of Computational Methods in Engineering",
issn = "1134-3060",
publisher = "Springer Netherlands",
number = "3",

}

TY - JOUR

T1 - Mitigating Error and Uncertainty in Partitioned Analysis

T2 - A Review of Verification, Calibration and Validation Methods for Coupled Simulations

AU - Stevens, Garrison

AU - Atamturktur, Sez

PY - 2017/7/1

Y1 - 2017/7/1

N2 - Partitioned analysis involves coupling of constituent models that resolve different scales or physics by allowing them to exchange inputs and outputs in an iterative manner. Through partitioning, simulations of complex physical systems are becoming evermore present in the scientific modeling community, making the Verification and Validation (V&V) of partitioned models to quantifying the predictive capability of their simulations increasingly important. Partitioning presents unique challenges, as well as opportunities, for the V&V community. Verification gains a new level of complexity in partitioned models, as numerical errors can easily be introduced at the coupling interface where non-matching domains and models are integrated together. For validation, partitioned analysis allows the quantification of the uncertainties and errors in constituent models through comparison against separate-effect experiments conducted in independent constituent domains. Such experimental validation is important as uncertainties and errors in the predictions of constituents can be transferred across their interfaces, either compensating for each other or accumulating during iterative coupling operations. This paper reviews published literature on methods for assessing and improving the predictive capability of strongly coupled models of physical and engineering systems with an emphasis on advancements made in the last decade.

AB - Partitioned analysis involves coupling of constituent models that resolve different scales or physics by allowing them to exchange inputs and outputs in an iterative manner. Through partitioning, simulations of complex physical systems are becoming evermore present in the scientific modeling community, making the Verification and Validation (V&V) of partitioned models to quantifying the predictive capability of their simulations increasingly important. Partitioning presents unique challenges, as well as opportunities, for the V&V community. Verification gains a new level of complexity in partitioned models, as numerical errors can easily be introduced at the coupling interface where non-matching domains and models are integrated together. For validation, partitioned analysis allows the quantification of the uncertainties and errors in constituent models through comparison against separate-effect experiments conducted in independent constituent domains. Such experimental validation is important as uncertainties and errors in the predictions of constituents can be transferred across their interfaces, either compensating for each other or accumulating during iterative coupling operations. This paper reviews published literature on methods for assessing and improving the predictive capability of strongly coupled models of physical and engineering systems with an emphasis on advancements made in the last decade.

UR - http://www.scopus.com/inward/record.url?scp=84966459683&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84966459683&partnerID=8YFLogxK

U2 - 10.1007/s11831-016-9177-0

DO - 10.1007/s11831-016-9177-0

M3 - Article

AN - SCOPUS:84966459683

VL - 24

SP - 557

EP - 571

JO - Archives of Computational Methods in Engineering

JF - Archives of Computational Methods in Engineering

SN - 1134-3060

IS - 3

ER -