Hi, in some transient solution cases in Fluent , after thausands of time steps , not limited , the updating of solution time step to start the iteration of the next time step take a long time, some times reached 2 minutes , and in this period the program seems to be blocked and then resuming the calculation , so that the solution to the steady state conditions will take many hours, can anyone help me to solve this issue ?
Dear Alaa,
firstly you need to estimate the time scale of the process. What type of the process are you investigating ? The modelling process may have stiffness, so current time step is in depence of stability of the numerical scheme, but not on accounting error.
With best regards,
Serge.
Dear Alaa,
Do you mean there is some lapse between two consecutive time steps?
or it is continuously iterating but takes longer to solve later time step?
Hi,
which version of ANSYS Fluent are you using and have you tried a different or the latest version (19.2). Is the observation then still persisting?
Are you solving in serial, local parallel (shared memory) or in distributed parallel? on how many cores? Is the phenomenon dependent on the number of used cores?
Further I'm a bit confused. Why you are using transient simulation, if the aim of your simulation is to reach the steady state condition? Maybe this part of your question is a bit misleading or difficult to understand. For reaching steady-state condition as fast as possible I would use the steady-state, pseudotransient solution approach with an as large as possible pseudo-timescale (or auto-timescale). This would be much faster, since this solution approach doesn't make inner iterations for pseudo-timesteps (in contrary to a true transient solution). So maybe you can explain, why you solve your application this way? Maybe you are interested in the transient evolution of the flow as well until it reaches steady state. Then it's ok.
Regards,
Thomas.
With consideration to Thomas Frank about why do you conduct transient solution if you ask for steady state solution, I would recommend you to change (enlarge) the mesh size.
Just in case you write lots of monitored data to disk, this could slow down computation speed after a while when the files begin to get quite large. At least this happened to me once.
Dear Mr Serge Polyakov , I'm steady transient case of heat pipe , i.e. 2phase + vof model .
Mr. Thomas Frank I'm trying use Pseudo transient technique both fixed and variable time step, but the solution always go divergence!!
Dear Mr Mohamed Shedid , enlarge of mesh size lead the solution to the divergence in my case.
Ok, you get the point Mr Julian Vogel , what you do to solve this issue?
Unfortunately, in my case, the only way was to reduce the amount of monitors or the write frequency.
Yes, Mr Julian Vogel your opinion was write , I'm reduce it to minimum but the elapsed time still over a minute.
- Badges
- Science topic