no mass convergence
 

I have a problem which I am sure is so simple that when I find the solution I will be embarrassed. I have a device that I want to place in a channel. I built a mesh of the device and ran it using STAR-CD. I had no problems and the solution converged with residuals to 10-5. I then extruded an inlet and exit section, and performed vmerge. A connectivity check shows only one region. When I try to solve this, the mass does not converge. If I check mass flow rate after 1600 iterations, the inlet is as I set it, but before my structure in the channel it is down to almost nothing. If I remove the inlet and run with the outlet portion, by the time I exit my structure I am again down to no flow. Obviously, my inlet does not see the outlet. Therefore I have tried the following: 1. merge all vertices â€" no effect 2. remake all cp matches â€" no effect 3. use default wall friction â€" no effect 4. decrease inlet flow â€" no effect 5. qhid plot shows everything as one volume 6. connectivity check shows ones volume

So, what now?

Thanks, Tom

Re: no mass convergence
 

Hi Tom,

It seems, we have the same problem:

http://www.cfd-online.com/Forum/starcd.cgi?read=9218

In V3,15 we fixed the problem by putting the default walls (region 0) to a seperate region. In V3.24 we do not have any idea, what to do.

CU

matze

Re: no mass convergence
 

I sent my problem to adapco tech support. They made no changes to the mesh. Instead they:

- AMG with rcon55=1 (RCON because running LINUX)

- UVW mom relaxation factor = 0.5, residual tol=0.01

- Pressure RF = 0.01, Red tol=0.001

- Viscosity = 0.95

- SW100 ON

- Turned ON conservation checks and input data check from Analysis Controls > Output controls > Monitor Numeric Behavior

I am now on iteration 1275 and the mass residual is 5e-04 and still decreasing.

I still don't understand, but maybe this will help you. Let me know.

Tom

Re: no mass convergence
 

I too have the same problem regarding mass convergence. initially i felt the mesh was the problem, however after varying the mesh several times i eliminated this possibility.

i am modelling a very basic problem, after viewing the info file, i found that there is problem with the outlet boundary which forces the mass residual to zero. viewing the results, the flow seems to be bouncing off the outlet in certain regions, however this may be due to recirculation in the domain

any ideas on what can be done to fix this?

i have tried playing around with the boundary conditions and relaxation factors etc

Re: no mass convergence
 

Try using the AMG solver!!!. I had the same problem before, and had been solved using the AMG solver, I do not know why. Best wishes with fast convergence. Wael

no mass convergence
 

thanks for your response Wael. I tried the AMG solver, it didnt work!

im trying to solve a laminar flow problem here, if that makes any difference. UVW all convergence within about 10-20 iterations, however, Mass residual remains constant for about 10 iter then it usually diverges or fluctuates.

Re: no mass convergence
 

nice, this funny switches... amg was a good way to fix the problem: abort with error message in first iteration! CD-Adapco told us to check, if all cells are connected in one region and to do a crack check. I dont know, how many days this crack check will need :( One of two problem models nox is fixed by eleminating the second region.

matze

Re: no mass convergence
 

What means "AMG didnt work" ? Do you get an abort with error message? If yes, check your grid (all connected, no cracks, all couples correct?). We had two seperate regions (the second was very small...)

matze

Re: no mass convergence
 

I tried the AMG solver to try and get the mass residual to converge instead of diverge, the solver does run but the mass residual continues to diverge.

i have checked cracks, couples, vertices etc (via the check everything function).

how did u find the very small separate region?

Re: no mass convergence
 

cell tool > check > connectivity

you get "results" in register 4, each region is one number. To see special regions, try cset news prange 4 <from_value> <to_value>

matze