[Situla]

Hello everyone!


I see many questions about the residuals, but very few answers. I personally am interested in what which one is. In my history.vtk I have the following columns (among others)
"Res_Flow[0]","Res_Flow[1]","Res_Flow[2]","Res_Flow[3]","Res_Flow[4]","Res_Turb[0]"
In the FAQ I found answers to things called "Conservative", "Residual", and "Primitive". How are they related to the ones in the history file? What are "Res[nu]" and others?

The link mentioned in this post does not seem to exist; it just leads to the github page.



Another thing: Sometimes I get a lower value for "log10[Maximum residual]" in the Residual Evolution Summary than some of the residuals of the preceding or succeeding iteration (e.g. -8 for the "log10[Maximum residual]" but -6 for "Res[Rho]" or some other). How is that possible? Logically, I would assume that SU2 calculates all residuals in all cells, then for each variable (e.g. density, x-velocity) finds the maximum of all cells, then it writes those in the command output as well as in history file, and then it finds the maximum of the maxima of the variables and lists that as "log10[Maximum residual]". Clearly, this is not the case, but I am struggling to figure out what else it can mean, and I cannot seem to find the answer in the code without spending hours trying to understand it.



Thank you for your time.

[wallym]

Hi Situla,


Thanks for using SU2 and coming to the forum!


To answer your question the Res_Flow[#] correspond to residuals of the conservative variables. So in the 2D Euler case, we have rho = Res_Flow[0] rho*u=Res_Flow[1] and so on.
So in this case the Res[Rho] and Res[rhoE] should match up directly to Res_Flow[0] and [4] directly. Res[nu] = Res_Turb[0].



The second part of your question is a bit strange. Your logic is correct. The residuals written on the screen and history are the root-mean-square of the residuals of that variable for the entire geometry/mesh.

But, that being said the Log10[maximum] should be bigger than any of the output at that iteration, unless the residuals increase in the next iteration.


I hope that made sense. Please let me know if you have anymore questions.


Wally

[Situla]

Hi Wally! Thank you very much for the clarification. That helped a lot.


Regarding the second part, I am quite certain that I encountered this, but it might have been during some of the many unsuccessful attempts I performed in trying to get a simulation to converge. If I see it again, I will attach a screenshot.


Many thanks and regards

[Situla]

Hello!

To follow up on the uncertainty of the "maximum" not actually being the maximum of the residuals (or at least not of all residuals), the attached screenshot shows that the alleged maximum is clearly lower than the nu residual of the preceding and succeeding iterations.
Quite possibly, turbulent residuals are simply an exception and are not considered for evaluating the maximum.

Kind regards

[akoodlyr]

Hi,

IIRC the maximum residual that gets printed is the maximum among the flow solver residuals which would explain the turbulent solver residual being larger than the one printed. SU2 solves the flow and turbulent equations in a segeregated manner and, as you said, the output file only looks at the maximum among the flow solver residuals and not all equations.

I hope this resolves some confusion.

Akshay