Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
time dependent solver vs. time step vs. initial step
Posted 26 lug 2012, 11:06 GMT-4 Studies & Solvers Version 4.3 4 Replies
Please login with a confirmed email address before reporting spam
Hi
I am simulating a coupled model of fluid flow and heat transfer in fluid, range(0,2592000,946080000). Despite that the time dependent solver runs infinity, I have been trying out different workarounds:
- shorten the range to range(0,86400,946080000)
- initial time step: 0.000864
- relative tolerence: 0.1, 0.01; absolute tolerence: 0.01, 0.001
- time dependent solver BDF: strict, intermediate
nothing works, my time dependent simulation still runs forever without returning any results.
I expend 'solver configuration' and try to 'compute to selected' to dependent variables, it shows the initial values, but once I compute to 'time dependent solver', it runs forever again.
Please help me! I have been working on this issue for almost two weeks...
ps: I have got all the initial values/BCs from a stationary run.
Thanks!!
regards
Liwah
I am simulating a coupled model of fluid flow and heat transfer in fluid, range(0,2592000,946080000). Despite that the time dependent solver runs infinity, I have been trying out different workarounds:
- shorten the range to range(0,86400,946080000)
- initial time step: 0.000864
- relative tolerence: 0.1, 0.01; absolute tolerence: 0.01, 0.001
- time dependent solver BDF: strict, intermediate
nothing works, my time dependent simulation still runs forever without returning any results.
I expend 'solver configuration' and try to 'compute to selected' to dependent variables, it shows the initial values, but once I compute to 'time dependent solver', it runs forever again.
Please help me! I have been working on this issue for almost two weeks...
ps: I have got all the initial values/BCs from a stationary run.
Thanks!!
regards
Liwah
Attachments:
4 Replies Last Post 27 lug 2012, 06:48 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I'm not by my WS so I cannot have a look at your model, but I suspect that for t=0 your initial conditions are too far from a solution. Have your tried to define first a steady state solver to get a coherent flow& T distribution at t=0 and use that as initial conditions for your time series ?
--
Good luck
Ivar
I'm not by my WS so I cannot have a look at your model, but I suspect that for t=0 your initial conditions are too far from a solution. Have your tried to define first a steady state solver to get a coherent flow& T distribution at t=0 and use that as initial conditions for your time series ?
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar
I ran a steady state to get the initial temperature, pressure etc, to be assigned into the subsequent transient state. Previously with 4.2a, I successfully ran a similar transient state (with less BCs and some unrealistic parameters), but it got kind of crashed when I imported to 4.3, therein I rebuild a new one, run a stationary state, then a transient state. Stationary state returned good result but when the transient state is running infinity...
On the other hand, I am wondering if I should modify anything at the solver i.e. segregated step, initial time step etc..
Thanks for your time, Ivar!
regards
Liwah
I ran a steady state to get the initial temperature, pressure etc, to be assigned into the subsequent transient state. Previously with 4.2a, I successfully ran a similar transient state (with less BCs and some unrealistic parameters), but it got kind of crashed when I imported to 4.3, therein I rebuild a new one, run a stationary state, then a transient state. Stationary state returned good result but when the transient state is running infinity...
On the other hand, I am wondering if I should modify anything at the solver i.e. segregated step, initial time step etc..
Thanks for your time, Ivar!
regards
Liwah
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I too rebuild my models, as I do not fully trust the translations, COMSOL is still evolving quite a lot between versions, as new improved features arrive, and its rather quick to rebuild a model. What you are saying confirms what I read in the release notes, that the solver settings have somewhat been adapted. But I have not had enough free time to test this out yet in v4.3.
If you model is not too large, I can only suggest to add a "fully coupled" node to the solver sequence as for me tis often solves quicker, but requires more RAM. It' then only to disable the fully coupled npode to revert back to COMSOLs default settings. Pls note that once you manually add or adapr the solver sequence, COMSOl flags the sequence as user defined, and I have noticed that it does not lways adapt the solver sequence automatically , if you change elements and nodes in your model, so you should check t elogic manually after modifying your model.
Another point is the order of the physics and the segregated solver steps, sometimes the default order of solving the dependent variables is not optimum, and it's worth to reorganise (up/down) the order in which you solve the variables. This requires manual tweaking and your knowledge of the process going on: which variable is the "master, which one the "slave", even if they are coupled.
--
Good luck
Ivar
I too rebuild my models, as I do not fully trust the translations, COMSOL is still evolving quite a lot between versions, as new improved features arrive, and its rather quick to rebuild a model. What you are saying confirms what I read in the release notes, that the solver settings have somewhat been adapted. But I have not had enough free time to test this out yet in v4.3.
If you model is not too large, I can only suggest to add a "fully coupled" node to the solver sequence as for me tis often solves quicker, but requires more RAM. It' then only to disable the fully coupled npode to revert back to COMSOLs default settings. Pls note that once you manually add or adapr the solver sequence, COMSOl flags the sequence as user defined, and I have noticed that it does not lways adapt the solver sequence automatically , if you change elements and nodes in your model, so you should check t elogic manually after modifying your model.
Another point is the order of the physics and the segregated solver steps, sometimes the default order of solving the dependent variables is not optimum, and it's worth to reorganise (up/down) the order in which you solve the variables. This requires manual tweaking and your knowledge of the process going on: which variable is the "master, which one the "slave", even if they are coupled.
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi Ivar
many thanks for the advise.
I added a 'fully coupled' node, the 'segregated' node was greyed out by itself then. I moved 'fully coupled' node a level up (above 'iterative' node), run! the computation takes less RAM in my case (<= 2GB, if I use 'segregated' node, it takes <=10GB). It is still running and I am not sure if it will turn out to be another infinity.
May I check with you if I should stick with 'iterative' or 'direct' under 'time dependent solver'? I read some threads saying that 'iterative' runs faster than 'direct', but some threads say the otherwise. I am pretty confused.
I shall keep it posted! Many thanks again, Ivar!
regards
Liwah
many thanks for the advise.
I added a 'fully coupled' node, the 'segregated' node was greyed out by itself then. I moved 'fully coupled' node a level up (above 'iterative' node), run! the computation takes less RAM in my case (<= 2GB, if I use 'segregated' node, it takes <=10GB). It is still running and I am not sure if it will turn out to be another infinity.
May I check with you if I should stick with 'iterative' or 'direct' under 'time dependent solver'? I read some threads saying that 'iterative' runs faster than 'direct', but some threads say the otherwise. I am pretty confused.
I shall keep it posted! Many thanks again, Ivar!
regards
Liwah