No convergence issue
Hello everybody
I am trying to run a simulation of a reactor with the sst k omega, however I have some problems with convergence. I use the simpleFoam solver with 12 processors. fvSolution Code:
/*--------------------------------*- C++ -*----------------------------------*\ Code:
/*--------------------------------*- C++ -*----------------------------------*\ |
1 Attachment(s)
U
Code:
outlet Code:
boundaryField Code:
/*--------------------------------*- C++ -*----------------------------------*\ Code:
CMA: no RDMA devices found the geometry consist on a pipe with many holes on its wall, releasing air in the main core of the reactor which is cylindrical can somebody help me on that? thank you all |
2 Attachment(s)
I attach also some slice where you can see the main pipe on the left and the differents holes of injection
|
Hi,
1) Can you make a better mesh without these skewed cells? 2) I don't really understand the geometry. Maybe it is not important, but could you try to post a picture of the whole thing, with highlighted inlets and outlets? 3) Can you post some log output? 4) In your residual plot you see some convergence finally at the end. Why did you stop calculating and what did you change at it=12000? |
As I understand it your solver settings are not very good. If there is no special reason you should not deactivate the relTol in SIMPLE. In SIMPLE it is not needed that every single linear equation is solved to a very low residual (what you do by just using the "tolerance"). This is, because you do the outer (SIMPLE-) iterations anyway and you just need the inner interations solved in a way that the outer algorithm remains stable. Thus, I usually use relTol=0.1 and set the tolerance to some very low value, such as 1e-18.
Secondly, you set the laplacian scheme to "limited 1", which means you use the corrected scheme. This is not the most stable one. Begin with "limited 0", this is the uncorrected scheme and if that works you can start to increase the numer to 0.3, 0.5, ... |
thank you very much I will try your suggestions, i noticed there was something not right about the solvers. I will post here my results.
The problem is I do this mesh with snappy because is a really complicated mesh. Sorry I cannot post the entire mesh however the inlet is on the pipe on the left. Sorry I don't really remember what I've changed. I will post the log output when I go back to my office tomorrow. however thank you very much for your help, is really appreciated |
Hello, Kurdt89,
Im not very familar with the your simulation. But looks like you are simulating a reactor(is there bubbles of droplets?). As we know SST k-omega is highly appropriate for strong adverse pressure gradient. In my simulations, such as stirred tanks, bubble columns, I just use k-epsilon. Could you please share some experience with SST k-omega model in your reactor simulaiton? Thanks. |
In my case the problem is that I cannot control the global value of y+ inside the whole thing and also that different regimes exhist, there are laminar regions in which the k-epsilon family would certainly crush (already tested).
So i use k omega sst with scalable functions, which is fine but highly unstable, I still have problems with convergence. |
All times are GMT -4. The time now is 00:33. |