3.6 - Errors and Uncertainty, Verification and Validation
3.6 - Errors and Uncertainty, Verification and Validation
3.6 - Errors and Uncertainty, Verification and Validation
(2012)
“The consequences of inaccurate CFD results are at best wasted time, money and
efforts, and at worst catastrophic failure of components, structures or machines.”
(Versteeg & Malalasekera 2007)
Definitions of error and uncertainty (AIAA 1998; Oberkampf & Trucano 2002; Versteeg
& Malalasekera 2007):
Error: a recognizable deficiency in a CFD model that is not caused by lack of knowledge.
Uncertainty: a potential deficiency in a CFD model that is caused by lack of knowledge.
Definitions of error and uncertainty (AIAA 1998; Oberkampf & Trucano 2002; Versteeg
& Malalasekera 2007):
Error: a recognizable deficiency in a CFD model that is not caused by lack of knowledge.
Causes or errors are:
1. Numerical errors
1. Round-off errors
2. Iterative convergence errors
3. Discretization errors
2. Coding errors: mistakes or “bugs” in the software
3. User errors: human errors due to incorrect use of the software
Definitions of error and uncertainty (AIAA 1998; Oberkampf & Trucano 2002; Versteeg
& Malalasekera 2007):
( Φ ) Φ +
Φ + ( Φ ) 0
Iterative convergence errors:
• The final solution (after an infinite number of iterations) will satisfy:
Φ + ( Φ ) 0
• However, after “k” iterations, the final solution will not yet have been attained, so:
Φ + ( Φ ) 0
Iterative convergence errors:
• The difference is called the local residual (for grid cell i):
Φ Φ + ( Φ )
• The global residual is the sum of the local residuals over all M grid cells:
Φ Φ + ( Φ )
Iterative convergence errors:
• Residual is often normalized: scaled residual.
• Be careful with pre-set residual thresholds / limits in CFD codes: these are often not
sufficiently stringent! Often, it is needed to reduce these limits.
Discretization errors:
• Control of the magnitude and distribution of discretization errors by careful and high-
quality grid design is a major concern in CFD.
• Requires care, time and ingenuity to balance high-quality and low discretization
errors on the one hand and feasible grid size on the other hand.
Verification:
• The process of determining that a model implementation accuracy represents the
developer’s conceptual description of the model and the solution to the model.
• “Solving the equations right” (Roache 1998)
• Quantifying the errors
Validation:
• The process of determining the degree to which a model is an accurate representation
of the real world from the perspective of the intended uses of the model.
• “Solving the right equations” (Roache 1998)
• Quantifying the uncertainty
Verification and validation (AIAA 1998; Oberkampf & Trucano 2002; Versteeg &
Malalasekera 2007):
Verification:
• Estimation of discretization error: systematic refinement of grid and time steps with
monotonic reduction of discretization error on at least two but preferably three
successive levels of grid refinements.
Validation:
• Estimation of input uncertainty: by sensitivity analysis and uncertainty analysis
• Estimation of physical modeling uncertainty: by comparison of CFD results with
high-quality experimental results
Oberkampf & Trucano (2002): meaningful validation is only possible after good
quantitative estimates have been obtained of:
(i) All numerical errors
(ii) Input uncertainty
(iii) Uncertainty of experimental data used for validation
MODULE QUESTION
Which statement is correct?
A) Errors are assessed by verification, and uncertainties by validation
B) Errors are assessed by validation, and uncertainties by verification
C) Both errors and uncertainties are assessed by verification
D) Both errors and uncertainties are assessed by validation
In this module, we have learned about:
• Errors and uncertainty in CFD
• Verification and validation
In the next module, we will focus on:
• Best practice guidelines
Image: Blocken et al. (2012)