"All models are wrong but some are useful."
This citiation of George Box is also applicable to numerical simulations.
All simulations are approximations of the reality using numerical models. These models contain errors to a certain extent. Nevertheless with careful and serious work it is possible to perform simulations leading to useful and good results.
Today it is possible to perform simulations of physical phenomena without deeper knowledge of physics and numerics. Unfortunately, the unknown errors in such a simulation are not insignificant. Whether and in what extent the performed simulation matches reality stands in the background or is even unknown. Experience showed us that there shouldn't be blind trust to any software (at least) for calculating Fluid Dynamics or Fluid-Structure Interaction problems. There should be a lot of expertise in fluid dynamics, numerics and the used software package as a basic requirement to perform reliable simulations.
Therefore the need for reliable methods for bluff body flow simulations motivates a current topic of research.
The topic contains three big issues:
- Verification
- Validation
- Quantification of Uncertainties
Verification
Verification ensures that a computer resp. the used software correctly solves the given equations. Therefore the main concern of Verification is to prove the correctness of the solution of the pure mathematical problem without dealing with any real physics.
The topic of Verification can be divided into two issues. On the one hand, there is Code Verification and on the other hand there is Solution Verification.
Code Verification ensures that the software represents the given mathematical model exactly and that this model is solved correctly. For this purpose analytical problems are preferable because the correct (analytical) solution can directly be obtained.
Solution Verification ensures that the mathematical model used in a simulation works adequate for the intended use.
Validation
Validation represents a quantitative assessment of performed simulations using experimental measurements. These measurements are usually performed in a wind tunnel. Here, it is very important that simulation and experiment are performed in a strictly separated manner. This is absolutely necessary to guarantee that simulations can get predictive character.
Using this procedure it is possible to quantify the quality of the software for the intended area of application.
Quantification of Uncertainties
Uncertainties and scattering of parameters are a major challenge in many simulations.
In order to obtain reliable simulations and results it is necessary to find a procedure representing this stochastic behaviour in simulations that (actually) are deterministic.
Using methods of representative sampling could be a way to reply to this challenge.