Convergence

[xingjian]

Hello Luciano,

Thank you for your response, in the ‘Monitoring points output’ file format should be changed to ‘.csv’ so that the files may be read in ‘ParaView’ of Salome. But how many points do I need to choose? So that the results can be representative.

And my listing is a txt file, how can I run these code , In other words, how and where can I enter these codes in order to implement it?

Thanks very much

Kinds Regards

XING Jian

[Luciano Garelli]

Hello,

Yes, if you want to read in Paravis you have to save in csv format and about how many points, this will depend of your physical problem, the domain, etc. You have to know where to locate them.

The lines of code that I posted are not exclusive for CS so you don't have to implement anything, you have to execute in a terminal (I assume that you are using linux), inside of the result directory.

Regards,

Luciano

[xingjian]

Hello Luciano,

Thank you, you are so nice, but I am using Windows... So how can I do that?

Because in Fluent, I can directly observe the value of the residual is stable, to determine whether the results converge.

Is the result converging in Code Saturne? More specifically, is it to compare the value of Derive (which is in listing) and the value of Solver Precision(which is setted by me in Equation parameters of Numerical parameters)?

Thanks very much

Kinds Regards

XING Jian

[Yvan Fournier]

Hello,

The solver precision is used for incremental loops, possibly with several sweeps per time step, so that won't give you the most direct info about convergence.
Derive is much more useful for this. In v5.0, Derive is also in a .csv file, but I have not generated a Windows build for 5.0 beta yet (I'll release it when 5.0 is released, expected at the end of this month). The convergence tracker should also be runnable from the GUI. Otherwise, you can open csv files with other tools (ParaView, Excel, ...).

Using probes is also recommended, because when you have small fluctuations, Derive never geos very low, but you will see the time history stabilize. The trick is using a good sampling of probes placed in different locations.

Best regards,

Yvan

[xingjian]

Hello Yvan,

Thank you very much Yvan, you are so nice, because that I thought 'Relaxation of pressure increase' is incremental loops, and 'Solver precision' is Monitors Residuals. So, you mean that I do not need to compare the value of Derive and the value of Solver precision, as long as Derive is stable, no matter how big it is (The Derive of omega is even 0.58E+12), as long as it does not change, we can judge that the result is convergent?

By using probes, you mean that I can be able to choose some of the points that are characteristic, observing their speed, pressure, temperature, etc., if their changes are particularly small, we can judge the results are convergent?

And how can we judge the conservation of mass between inlet and outlet?

Thanks very much

Best Regards

Jian

[Luciano Garelli]

Hello,

In the listing file you can check the mass flow conservation in the section "BOUNDARY MASS FLOW INFORMATION"; there you have for each boundary condition type the mass flow and I think with a value of 0.58E+12 your solution it is not converging...

Regards,

Luciano

[xingjian]

Hello Luciano,

Thank you Luciano. So is N_iter number of iterations used to solve the linear system?
Norm. residual is normed residual after solving the linear system (should be lower than Soler Precision=1.e-8 which is the criterion for stopping the iterations)?
And does Derive mean |u(n+1)-u(n)|/|u|? The closer the value of Derive is to 0, the more convergent the result is？

In the section "BOUNDARY MASS FLOW INFORMATION", how to check the mass flow conservation? Can we compare |Mass flow of Inlet - Mass flow of Free outlet|/Mass flow of Inlet with 1%? For checking it's conservation or not.

Thanks very much

Best Regards

Jian

[xingjian]

Hello,

I really want to know that am i right?

Thanks very much

Best Regards

XING Jian

[Luciano Garelli]

Hello,

Yes, N_iter are number of iterations used to solve the linear system to the relative precision (epsilo).
You have a section in the user guide (8.2.11) about error estimators for Navier-Stokes. If the derive has dropped several order of magnitudes this means that the flow is not changing and you can estimate the mass flow conservation with the equation that you mention.

Regards

Luciano

[QuentinG]

Hello,

I just come back to ask a new question (and not create another topic with the same title), I am wondering what is the Code_Saturne equivalent to the Fluent residuals ?
Is it the Rhs norm the Norm residual or the derive ?

Thank you,

Regards,

Quentin