Factors of Safety for Richardson Extrapolation

Abstract

A factor of safety (FS) method for quantitative estimates of grid/time uncertainties for CFD solutions is derived to remove the deficiencies of GCI, corrected GCI(sub C), and correction factor (CF) methods, i.e., unreasonably small uncertainty when CF > 1 (estimated order of accuracy greater than theoretical) and lack of statistical analysis to prove 95% confidence for the estimated uncertainties to bound the true error. The approach follows the CF method but reflects the uncertainty instead of FS for CF < 1 for CF > 1 (CF = 1 is asymptotic range). FS at CF = 0 and 1 are determined by reliability and lower band of the confidence interval of the true mean based on statistical analysis using a large sample of analytical/numerical benchmarks covering 17 studies, 96 variables and 304 individual grid triplets. Only the FS method provides 95% confidence that the actual factor of safety FS(sub A) > 1 for the 304 grid convergence studies: confidence intervals are 86.2%, 92.1%, 91.5%, and 95.7% for GCI, GCI(sub C), CF, and FS. For 20% of the data 1.1 smaller or equal CF < 2.0, GCI, GCI(sub C), and CF methods fail as only 47.4%, 71.2%, 72.9% confidence intervals are achieved, whereas 89.3% is achieved for the FS method. Only the FS method has 95% confidence the lower band of the confidence interval for FS(sub A) is larger than 1.2 for different studies, variables, ranges of CF, and single CF values where multiple FS(sub A) are available.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2009
Accession Number
ADA498086

Entities

People

  • Fred Stern
  • Tao Xing

Organizations

  • University of Iowa

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Accuracy
  • Bending Stress
  • Boundaries
  • Coefficients
  • Computational Fluid Dynamics
  • Computational Science
  • Convection
  • Engineering
  • Equations
  • Extrapolation
  • Radial Velocity
  • Statistical Analysis
  • Statistical Distributions
  • Statistics
  • Three Dimensional
  • Transport Properties
  • Two Dimensional

Readers

  • Computational Modeling and Simulation
  • Hydraulic Engineering.
  • Statistical inference.