Discussion Forum

To add some more details I´ve taken a snapshot of the calculation. As you can see, the convergence criteria is satified at the last 3 steps but COMSOL just doesn´t stop. And if I stop the calculations manually, no result is given.

I really don´t know why this is happening or what I can do to resolve this issue.

Any help is appreciated..

Hi,

It would be easier to analyze you would append a screenshot of the solver log.

One guess however:

For the damped Newton solver to be converged, two things are required:

1. Convergence criteria fulfilled
2. No remaining damping of the Newton step ("Damping" in the solver log = 1)

So maybe criterion 2 is not fulfilled.

Regards,
Henrik

Thanks for your answer.

The log of the calculation is:

Anzahl der Freiheitsgrade, nach denen gelöst wird: 658104
Nichtsymmetrische Matrix gefunden.
Skalierung für abhängige Variablen:
Geschwindigkeitsfeld (mod1.u): 0.12
Druck (mod1.p): 1
Iter ErrEst Damping Stepsize #Res #Jac #Sol LinIt LinErr LinRes
1 58 0.0100000 58 2 1 2 6 0.062 0.18
2 6 0.1000000 6.6 3 2 5 8 0.086 0.81
3 0.023 1.0000000 0.7 4 3 8 23 0.12 1
4 0.04 1.0000000 0.038 5 4 10 35 0.091 0.93
5 0.029 0.4808729 0.055 6 5 13 64 0.076 0.38
6 0.0049 1.0000000 0.031 7 6 16 87 0.027 1.1
7 0.0023 1.0000000 0.0049 8 7 19 103 0.024 1
8 0.0096 0.1024721 0.011 9 8 22 127 0.047 0.26
9 0.009 0.2100768 0.011 10 9 25 160 0.1 0.1
10 0.00092 1.0000000 0.0093 11 10 28 178 0.12 1.1
11 0.00035 1.0000000 0.0013 12 11 31 192 0.12 1
12 0.00058 0.4346526 0.001 13 12 34 212 0.11 0.44
13 0.00035 0.6217532 0.00089 14 13 37 241 0.045 0.61
14 9.1e-005 1.0000000 0.00051 15 14 40 267 0.1 1
15 8.3e-005 0.6829436 0.00025 16 15 43 297 0.052 0.88
16 5.1e-005 0.7353162 0.00019 17 16 46 322 0.06 0.95
17 4.6e-005 0.6494489 0.00013 18 17 49 349 0.051 0.97
18 2e-005 0.8071157 9.6e-005 19 18 52 375 0.074 0.99
19 3e-005 0.5304653 6.4e-005 20 19 55 404 0.039 0.99
20 5.6e-006 1.0000000 5e-005 21 20 58 430 0.11 1
21 0.00021 0.1000000 0.00023 22 21 61 466 0.036 0.79
22 4.5e-006 1.0000000 0.00021 23 22 64 488 0.061 1
23 0.00064 0.1000000 0.00071 24 23 67 525 0.047 0.15
24 1.2e-005 1.0000000 0.00065 25 24 70 547 0.12 1.1
25 3.2e-006 1.0000000 2.7e-005 26 25 73 573 0.043 1


As you can see there are some values that fulfill both criteria you mentioned (ít also shows "100%" at various points). So I dont see the reason why the solver does not stop computing...

Can you see any problems in the log?

best regards

Christian

This is a guess because I have seen similar behavior here.

Version 4.3b added the option to converge on either the solution or the residual.or both.

If you choose the solution, it appears to behave as it did prior to version 4.3b, and the convergence is satisfied based on what you can see in the log files.

However, if you choose the residual, then apparently the convergence criteria is more stringent, and it takes longer to satisfy the residual convergence. I don't think we can see the residual in the log file.

I like to pick "solution" just for this reason.

Hi

This problem can happen in using pseudo time stepping in solver

Best regards