Verification and Validation Plan

1
Verification and Validation Plan

• Michael Ortiz
• Caltech PSAAP Site Visit
• September 24-25, 2007

2
VV and UQ techniques
3
UQ and priorities
The Yearly Assessment Cycle
4
Verification Validation Plan

• Table of input parameters and performance
  measures, integrated experiments IE1-IE3

5
Verification Validation Plan

• Table of input parameters and performance
  measures, integrated experiments IE4-IE5

6
Verification Validation Plan

• Table of configurations, diagnostics and data
  acquired for component experiments CE1-CE3

7
Verification Validation Plan

• Table of validation milestones and data sources

8
UQ Caltech style

• Goal-oriented (certification) approach to UQ
• VV UQ A code is verified, validated, when
  associated uncertainties (aleatoric, epistemic)
  are small enough
• UA DF, UE DF-G rigorous quantitative
  measures of verification and validation
• When is enough enough? U UAUE small enough!
• Uncertainties (aleatoric, epistemic) in all
  performance measures will be computed
  (verification and validation diameters) yearly,
  reported at the end of the Yearly Assessment
  cycle, to track progress towards achieving a
  verified and validated predictive capability (cf.
  VV plan)

9
UQ challenges

• Open research questions Are concentration-of-meas
  ure inequalities tight enough to be useful
  towards certification?
• How can concentration-of-measure inequalities be
  improved upon?
• Correlated inputs
• General probability density functions for inputs
• UQ in a multiscale, hierarchical, multicomponent
  setting
• Use of archival data, limited ability to conduct
  integral tests
• Are uncertainties computable with present
  computational resources?
• How can UQ calculations be optimized, sped up?