Convergence/Residual calculation
 

Hello OpenFoamers,

I want to understand what the monitoring Data are to find the problem for the course of my convergencecurves (which is bad).

1) Can someone tell me please what exatctly the initial and final values are?
2) Which equation are the "residuals" from?
and another Question
3) Can I let me show the residuals for one vertice?

Thanks for answering so far!

Carolin

Hej

for your residual curves follow this small and easy tutorial by wolfgang http://www.cfd-online.com/Forums/ope...residuals.html

1. the initial value is the residual before the sub iterations start, meaning the change between the value of the previous iteration and the start of this iteration, the final value is the residual between the 0th iteration and the last iteration of an equation after the subiterations have been performed.
2. the residuals come from the respective equations for a variable, for example the p or p_rgh residual comes from the differential equation for the pressure (take a look at SIMPLE and PIMPLE algorithms to see why there is a pressure equation). it is easier to see for example the temperature equations, just look into the solver buoyantBoussinesqSimpleFoam and look at the file T, which gives you the equation for the temperature for which then a T residual will be computed.
3. I am not sure if you can get the residual for one vertice without rewriting the output. You could sample that value over and over again and see how it evolves over the iterations, from this you can see how the residual (final residual/value) changes. You wouldn't get the sub iteration residuals from there.

Hey Roman,

thanks for the fast reply! Ive looked into the solver equations, but im not really firm with c++. So I still dont understand:
1) Is the plotted value for init/final residual the average value of all solved equations in the mesh or the biggest value? Iam now just thinking of one iteration step, which gives one residual value for init and final for eg p/U/k etc. . Do you know that?

Miriam

The is only one equation per variable (U, p_rgh, T, k, epsilon, ...) solved (which is a matrix equations, so I think you could say that there are many equations). The residual value should be the largest value of the matrix, since averaging would hide the largest value and therefore would hide the convergence in all cells of the mesh.

As for the solvers, take a look at http://www.openfoam.com/docs/user/pr...e-OpenFOAM.php The equations are quite easy to under stand that way, because they are kind of written the same way as you would write them on paper.

Hello Foamers,

Is there a way of calculating the averages of residuals in OpenFoam?
I have a log file with residuals but i want to plot the average values of them.

You may help to know that way if exist.


Regards!