Discussion Forum

How can i know if it has any problems in the simulation?

Hi

the Log and Progress windows give you some clues, the Convergence plot window others, then you should turn on the "plot while solving" option in the main solver node (a "closed tab" you need to open it) but this is plotting for each time you reach your desired time step results, and finally you might turn on the stepping per solver iteration, deeper down in the solver tree, depends on the solver selected, there are one or two other positions where you find a plot while solving tab

The plot while solving must be selected BEFORE you launch your solver

--
Good luck
Ivar

Hi Ivar
Thanks for your reply

Is normal that the solution take all this time? or maybe something is wrong?


Jason

Log window show this many time:

Inconsistent pointwise unidirectional constraints found.
16 constraints are merged/removed.

I can see that the stepsize is very small.

Hi

well it's not the best sign, but rotary machinery is complex so it is often time consuming. I try often to run y complex models with very crude mesh to get them tested ina reasonable time, then only I turn on fine stepping and decrease the meshing size.

--
Good luck
Ivar

Hi Ivar

Thank you very much for your suggestion.

I have another doubt. In Time Dependent solver, in time stepping section, is preferable select FREE or STRICT in steps taken by solver list ?. What considerations i have to take to select the steps? Because, i see that the stepsize choose for the method is very small (in the order of 10^-4)

Hi

the stepping time settings is "dependent on the model".

The default "free" uses the change in parameters to adapt at will, then to gently increase, this causes often very small steps at start, particularly when you give poorly defined initial conditions far from "reality".

"Strict" uses the time steps you decide, but if you are too gross the solver might not manage to converge, particularly again if initial conditions are poor.

Then you also have "intermediate". The latter means that COMSOL solves at least for one case in between your provided steps.

Then do you have a diffusion problem or a standing wave problem (oscillatory BC). For the former the "free" is well adapted as the step taken increases (typically doubles) regularly when the solution tends exponentially to its asymptote, you gain solver time like this, but the risk is that if you pulse any BC late in the solving steps, the solver might skip fully your fine changes, and you get wrong results. Oscillating or turn on turn off conditions are typical subjects for strict or I find better intermediate solver settings. Such to ensure the solver catches all the oscillations of your BC's

--
Good luck
Ivar

Hi Ivar
I choose Generalized alpha method in time stepping, select Manual in steps taken by solver and range of time is (0,50,32400)

The problem es:

Nonlinear solver did not converge.
Time : 1425.000047683716
Last time step is not converged.


What should i do? I treat to reduce the tolerance and decrease the size of mesh

If is possible, please check the file.

Thank you very much for your help.


Jason

Hi Ivar

I hope u can help me, i still have the same problem