[braennstroem]

Hi,

we try to calculate a jet flow in a room using simpleFoam. Inlet velocity is set to 2.6m/s; for the outlet pressureoutlet is used (p=0) and 'wallFunctions' for the surrounding walls.

Unfortunately we have strange 'singularity problems' and somehow no 'visible' time steps from 3 to 44 in the log as shown below:

Starting time loop

Time = 1

BICCG: Solving for Ux, Initial residual = 1.99239e-05, Final residual = 8.52279e-08, No Iterations 1
BICCG: Solving for Uy, Initial residual = 0.000382404, Final residual = 1.11723e-06, No Iterations 1
BICCG: Solving for Uz, Initial residual = 0.00130972, Final residual = 3.30692e-06, No Iterations 1
ICCG: Solving for p, Initial residual = 1, Final residual = 0.00981217, No Iterations 76
time step continuity errors : sum local = 4.51616e-06, global = -2.50011e-07, cumulative = -2.50011e-07
BICCG: Solving for epsilon, Initial residual = 0.0527458, Final residual = 8.61077e-06, No Iterations 1
BICCG: Solving for k, Initial residual = 1, Final residual = 0.00244073, No Iterations 1
ExecutionTime = 65.93 s ClockTime = 66 s

Time = 2

BICCG: Solving for Ux, Initial residual = 1.62207e-05, Final residual = 4.69139e-08, No Iterations 1
BICCG: Solving for Uy, Initial residual = 0.000286247, Final residual = 9.1854e-07, No Iterations 1
BICCG: Solving for Uz, Initial residual = 0.00106412, Final residual = 3.19452e-06, No Iterations 1
lution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for Uz: solution singularity
ICCG: Solving for p: solution singularity
time step continuity errors : sum local = nan, global = nan, cumulative = nan
BICCG: Solving for epsilon: solution singularity
BICCG: Solving for k: solution singularity
ExecutionTime = 6989.99 s ClockTime = 7395 s

Time = 45

BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for Uz: solution singularity
ICCG: Solving for p: solution singularity
time step continuity errors : sum local = nan, global = nan, cumulative = nan
BICCG: Solving for epsilon: solution singularity
BICCG: Solving for k: solution singularity
ExecutionTime = 7307.35 s ClockTime = 7714 s


We used a one time step potential flow solution as initial conditions and reduced the relaxation factors to p=0.0001 and the rest to 0.2.

Do you have any suggestions what we can do?

Greetings!
Fabian

[gschaider]

No idea.

My suggestion:
Have you written the solution after the first time-step? Usually looking at that gives a good hint (crazy velocities etc). And try it without turbulence (can't think of anything els you can switch off, to isolate the problem)

The time warp looks real interesting.

[braennstroem]

Hi,

over night some strange things happened: the simpleFoam case runs without any solution singularity at time step 800, but as mentioned above we have such a time warp again, but now at two different time steps. One at:

Time = 25

BICCG: Solving for Ux, Initial residual = 1.02744e-05, Final residual = 2.61834e-08, No Iterations 1
BICCG: Solving for Uy, Initial residual = 0.000112404, Final residual = 3.61779e-07, No Iterations 1
BICCG: Solving for Uz, Initial residual = 0.000370707, Final residual = 1.23968e-06, No Iterations 1
ICCG: Solving for p, Initial residual = 0.991788, Final residual = 0.00943836, No Iterations 263
time step continuity errors : sum local = 1.61653e-05, global = 8.07791e-08, cumulative = -3.11282e-06
BICCG: Solving for epsilon, Initial residual = 0.000213944, Final residual = 1.55127e-07, No Iterations 1
BICCG: Solving for k, Initial residual = 0.00118659, Final residual = 3.08639e-06, No Iterations 1
ExecutionTime = 1800.61 s ClockTime = 1805 s

Time = 26

BICCG: Solving for Ux, Initial residual = 1.0157e-05, Final residual = 2.60278e-08, No Iterations 1
BICCG: Solving for Uy, Initial residual = 0.000110948, Final residual = 3.56346e-07, No Iterations 1
BICCG: Solving for Uz, Initial residual = 0.000366185, Final residual = 1.21793e-06, No Iterations 1
ICCG: 0678 s

Time = 54

BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for Uz: solution singularity
ICCG: Solving for p: solution singularity
time step continuity errors : sum local = nan, global = nan, cumulative = nan
BICCG: Solving for epsilon: solution singularity
BICCG: Solving for k: solution singularity
ExecutionTime = 10572.6 s ClockTime = 11076 s



And the second one back from 140 to 73:


Time = 139

BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for Uz: solution singularity
ICCG: Solving for p: solution singularity
time step continuity errors : sum local = nan, global = nan, cumulative = nan
BICCG: Solving for epsilon: solution singularity
BICCG: Solving for k: solution singularity
ExecutionTime = 64227 s ClockTime = 65789 s

Time = 140

BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
al = 0.000774369, Final residual = 2.21301e-06, No Iterations 1
ExecutionTime = 5199.55 s ClockTime = 5313 s

Time = 73

BICCG: Solving for Ux, Initial residual = 6.89053e-06, Final residual = 6.89053e-06, No Iterations 0
BICCG: Solving for Uy, Initial residual = 8.43502e-05, Final residual = 2.05703e-07, No Iterations 1
BICCG: Solving for Uz, Initial residual = 0.000290531, Final residual = 6.80189e-07, No Iterations 1
ICCG: Solving for p, Initial residual = 0.967161, Final residual = 0.00946104, No Iterations 267
time step continuity errors : sum local = 2.25708e-05, global = -1.6871e-07, cumulative = -2.90604e-06
BICCG: Solving for epsilon, Initial residual = 0.000100511, Final residual = 1.14051e-07, No Iterations 1
BICCG: Solving for k, Initial residual = 0.000770152, Final residual = 2.1998e-06, No Iterations 1
ExecutionTime = 5269.9 s ClockTime = 5383 s


Do you have an idea?

Greetings!
Fabian

[gschaider]

Just checking: You're using an umodified simpleFoam? You're not using any of the adaptive time-step stuff?

[hjasak]

Bernhard, there's no adaptive time step here - it's a steady state solver (time is only used as iteration counter).

This looks to me like a problem at the outlet or your solution has already gone bananas. Most of the stuff looks OK, but the initial pressure residual of (0.967161) is way too high for 70-odd iterations.

I would suggest looking carefully at your solution and discretisation: either there's something obviously wrong with the flow, the turbulence is all wrong or you're using central differencing on some convection terms you should not. In short, visual inspection required.

Hrv

[gschaider]

Hrv, I know that it's steady-state. The adaptive stuff was just a silly notion I had because I was to lazy to look at the source and I couldn't think of any other explanation for the time warps. Now I looked at the source and I have still no idea where the time-warps come from.

If the two runs are with identical settings and they're failing in different ways I have an even sillier idea: maybe it's not Foam. I know this is the lamest excuse in the book: maybe some RAM is failing (never witnessed that myself, but that indeterministic behaviour would be what I would expect in such a case; on the other hand why would Fabian's machine still run)

The only other explanation (I can think of, but I have a very limited imagination) for the time-warps would be that somebody is modifying the controlDict during the simulation (which I think is read during the runTime++) without Fabian's knowledge. Which is not a nice prank.

[steja]

I have a similar problem with turbFoam but only if I use OF1.3, i.e. running OF1.2 on the same setup everything is o.k.:

OF1.3:

Starting time loop

Time = 5e-05

Mean and max Courant Numbers = 0.0026444 0.117819
BICCG: Solving for Ux, Initial residual = 0.00605567, Final residual = 1.32513e-06, No Iterations 2
BICCG: Solving for Uy, Initial residual = 0.0218626, Final residual = 4.30722e-06, No Iterations 2
BICCG: Solving for Uz, Initial residual = 0.0224129, Final residual = 4.56185e-06, No Iterations 2
AMG: Solving for p, Initial residual = 0.974855, Final residual = 7.20409e-07, No Iterations 28
AMG: Solving for p, Initial residual = 0.00219321, Final residual = 9.15636e-07, No Iterations 6
time step continuity errors : sum local = 7.1722e-10, global = 1.12086e-11, cumulative = 1.12086e-11
AMG: Solving for p, Initial residual = 0.0295064, Final residual = 8.58992e-07, No Iterations 24
AMG: Solving for p, Initial residual = 0.000449758, Final residual = 9.81688e-07, No Iterations 5
time step continuity errors : sum local = 7.20766e-10, global = -3.60251e-12, cumulative = 7.60611e-12
BICCG: Solving for epsilon, Initial residual = 0.0109446, Final residual = 9.07372e-06, No Iterations 1
BICCG: Solving for k, Initial residual = 1, Final residual = 9.971e-07, No Iterations 3
ExecutionTime = 11.61 s ClockTime = 12 s

Time = 0.0001

Mean and max Courant Numbers = 0.00255982 0.120631
BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for Uz: solution singularity
AMG: Solving for p, Initial residual = nan, Final residual = nan, No Iterations 501
AMG: Solving for p, Initial residual = nan, Final residual = nan, No Iterations 501
time step continuity errors : sum local = nan, global = nan, cumulative = nan
AMG: Solving for p, Initial residual = nan, Final residual = nan, No Iterations 501
AMG: Solving for p, Initial residual = nan, Final residual = nan, No Iterations 501
time step continuity errors : sum local = nan, global = nan, cumulative = nan
BICCG: Solving for epsilon: solution singularity
BICCG: Solving for k: solution singularity
ExecutionTime = 157.94 s ClockTime = 159 s


With OF1.2 everything runs fine:

Starting time loop

Time = 5e-05

Mean and max Courant Numbers = 0.0026444 0.117819
BICCG: Solving for Ux, Initial residual = 0.00605567, Final residual = 1.32513e-06, No Iterations 2
BICCG: Solving for Uy, Initial residual = 0.0218626, Final residual = 4.30722e-06, No Iterations 2
BICCG: Solving for Uz, Initial residual = 0.0224129, Final residual = 4.56185e-06, No Iterations 2
AMG: Solving for p, Initial residual = 0.974855, Final residual = 5.6738e-07, No Iterations 33
AMG: Solving for p, Initial residual = 0.0021932, Final residual = 9.30452e-07, No Iterations 6
time step continuity errors : sum local = 7.28832e-10, global = 5.95128e-12, cumulative = 5.95128e-12
AMG: Solving for p, Initial residual = 0.0295062, Final residual = 9.88774e-07, No Iterations 20
AMG: Solving for p, Initial residual = 0.000449719, Final residual = 4.93959e-07, No Iterations 6
time step continuity errors : sum local = 3.62678e-10, global = -4.78015e-12, cumulative = 1.17113e-12
BICCG: Solving for epsilon, Initial residual = 1, Final residual = 7.44754e-10, No Iterations 1
BICCG: Solving for k, Initial residual = 1, Final residual = 1.01302e-06, No Iterations 3
ExecutionTime = 14.35 s


Time = 0.0001

Mean and max Courant Numbers = 0.00255982 0.120631
BICCG: Solving for Ux, Initial residual = 0.0157106, Final residual = 8.84689e-07, No Iterations 2
BICCG: Solving for Uy, Initial residual = 0.0641733, Final residual = 2.42521e-06, No Iterations 2
BICCG: Solving for Uz, Initial residual = 0.0655767, Final residual = 2.67305e-06, No Iterations 2
AMG: Solving for p, Initial residual = 0.790297, Final residual = 7.86614e-07, No Iterations 31
AMG: Solving for p, Initial residual = 0.00732114, Final residual = 6.37712e-07, No Iterations 11
time step continuity errors : sum local = 1.49292e-10, global = -5.40232e-12, cumulative = -4.23119e-12
AMG: Solving for p, Initial residual = 0.0683789, Final residual = 7.55717e-07, No Iterations 23
AMG: Solving for p, Initial residual = 0.0012786, Final residual = 6.3544e-07, No Iterations 7
time step continuity errors : sum local = 1.29184e-10, global = 7.67925e-12, cumulative = 3.44805e-12
BICCG: Solving for epsilon, Initial residual = 0.134013, Final residual = 3.11612e-11, No Iterations 1
BICCG: Solving for k, Initial residual = 0.158505, Final residual = 5.05454e-06, No Iterations 2
ExecutionTime = 25.99 s


I tried several modifications of relaxations factors, interpolation schemes (upwind) for grad and div schemes but nothing changed. Also using ICCG for pressure gives the same result. Any suggestions are welcome.

Thanks, Steffen

[gschaider]

Have you already inspected the solution after the first time-step (as already suggested twice in this thread)?

[braennstroem]

Hi,

thanks for the suggestion! The bc and ic were probably mixed up a bit; we started to set 'k' and 'epsilon' to 1 after we had singularity issues every run before. It seems to run without any time warps now, but due to the 'weird' bc and ic the jet looks kind of funny. We will start over ...

But I wonder that the bc and/or ic can influence the time step procedure!?

Greetings!
Fabian

[hjasak]

Of course they can: with k and epsilon = 1 you get the turbulen vicsosity of 0.09 compared with the laminar viscosity of e.g. 1-e6, i.e. you are out by 3 orders of magnitude. This would explain a lot...

Hrv

[braennstroem]

Hi Hrvoje,

I just thought, that writing out the actual time step is done in some kind of loop, which is not affected by the bc and ic at all.

Greetings!
Fabian

[hjasak]

Hi Fabian,

I see what you mean - I was much more interested in the NaNs bit than what happens afterwards. I'm worried by the fact that the solver blows up without any symptoms first so the only possible culprit is a weird-looking solution.

As for the time-step problem, this looks to me like a possible memory violation (someone over-writing bits of memory that contain the time information. Is this all out-of-box OpenFOAM stuff you are running? It woul be useful to set the environment variables to catch the error:


setenv FOAM_ABORT 1
setenv FOAM_SIGFPE 1
setenv FOAM_SETNAN 1


or to run the solver through valgrind and see what it says. I usually run the debug version with array bounds checking (good for me but not in general), which may tell you much more if you are developing your own code.

Hrv

[braennstroem]

Hi Hrvoje,

yes, it is out-of-box simpleFoam. We will try to set the environments and the debug version. I don't know valgrind yet, but I'll take a look at it; thanks!

Greetings!
Fabian

[hjasak]

You probably already have valgrind on your system and this is the easiets thing to try. Do

valgrind --tool=memcheck simpleFoam <root> <case>

and (with your permission) I'd like to have a look at the output.

A few more questions:
- Are you getting these outputs on the screen or into a log file?
- Is there any possibility that you've got two processes writing into a same log file at the same time and making a mess of it?
- Can you reproduce this with one of tutorial cases (just in case the library got corrupted somehow).
- Did you add anything into the FOAM, which required recompiling one of the libraries since you've installed it (maybe we've got an incomplete build here)

Hmm,

Hrv

Hello Foam users

i am trying to simulate Fluid Flow and Heat Transfer in a Mixing Elbow (a problem similar with the one in fluent tutorials).

the file can be downloaded MixingElbow.tar.gz

when i run the case i optain always:


BICCG: Solving for Ux: solution singularity
BICCG: Solving for Uy: solution singularity
BICCG: Solving for h: solution singularity
ICCG: Solving for pd: solution singularity
time step continuity errors : sum local = nan, global = nan, cumulative = nan
rho max/min : 0 0
ExecutionTime = 0.34 s ClockTime = 0 s


Can anybody have a look on my file and give me a hint?
thanks
Atzaru

sorry for posting in the wrong place, i will open a new thread.

atzaru

[braennstroem]

Hi Hrvoje,

I did write it out to a log file; I could send you a copy!? Writing from
two processes into the same log file is probably the best explanation,
but I do not think that writing it to a log file using the commandline
(simpleFoam <case> > logfile &) could mix it; it would rather erase the
old file and create a new one.

We did not add anything into version 1.3.

Thanks for the valgrind hint, it looks pretty easy!

Greetings!
Fabian