Learning Forum

[h395523899]

When my simulation did not compare experiments. How do I verify the accuracy of my simulation?

Currently I know to minimize the impact of the grid on the simulation and maintain grid independence. By looking at the flow of exits and entrances meets the quality conservation.

[José Mantovani]

Hello hsomenumber...

 For this you need create a simulation in the same way of the experimental study, modeling the geometry, boundary conditions, etc.. closing with the experimental data. Run your simulation and compare... If the numerical values converge, your simulation is good. If no, you need take a look of your mesh dependence, boundary conditions etc.. 

I hope I have helped you! The most annoying thing I believe would be converging the values of Cp and Cf so you should try various models of turbulence and a very good mesh next to the wall.

Have a good studies!

Regards,

Mantovani.

[h395523899]

Thank you for your answer

Regarding convergence, what kind of iterative curve is converged? The shape of the curve I often encounter is like a sound wave. Around a straight line (such as 10^-4) up and down equidistant fluctuations. If it converges, should the overall trend be downward?

[Raef.Kobeissi]

The curves you see in Fluent, for example, represents the residual error of the equations being solved, hence, the continuity, momentum and turbulence equations (if you are using RANS) the curves represent the amount of change between the solution in the last iteration and the solution in the iteration before. These values should decrease meaning the residual error is decreasing. Different cases have different convergence criteria. For example, you can create a residual curve for certain forces and try to study its progress with every iteration, when the residual error stabilises to a certain level you can then declare it converged, etc.