Problem with maximum number of Timesteps

Phantalor · July 8, 2013, 06:52

Hello and good morning,
I have set up a 2 Phase simulation (there is somewhere another post about the exact settingshttp://www.cfd-online.com/Forums/cfx...tml#post424961)
As ghorrocks suggested in that post, i changed the timestep settings now the 2 Phase simulation works as intended. but now there is a new problem.
I need to simulate not only the behavior until something happens but a timesweep of at least 5-10 seconds. The problem is because of the Timestepping settings for about 2 seconds of simulation cfx needs about 1e5 timesteps, and at this number of timesteps it exits with an error code (something like maximum number of timesteps reached) and also erases the data up to this point. (well maybe there is a backup but as my HDD crashed after the error it was erased becaus of linux HDD repair)

So here is my question, is there a way to heigthen the number of timesteps while using adaptive calculations or is there a way to run the simulation step by step (also 1e5 timesteps and then the next 1e5 etc.) and is there a way to write a batch that commands ansys to do so.

My Problem is that is that i do not see a way to clearly define it, because of the adaptive timestepping there is not really a way to know the time when it reaches 1e5 timesteps

Ps: I'm calculating with the Academic research package (maybe its dependend on the license)

ghorrocks · July 8, 2013, 09:21

I do not know for sure but I do not think an academic license limits the number of time steps.

I have a lot of experience in modelling MEMS microjet structures and to be honest I do not use CFX for it very much any more - it is far too slow. Have a look at Fluent. I have got Fluent running something like 20 times faster than CFX for micro drop ejection (that is free surface models with strong surface tension effects).

Phantalor · July 8, 2013, 09:41

Well i would like to use fluent if it brings better results in less time, but the Pc that is for me to use and the solvers on the other machines only have cfx installed and it seems we only have a licence for fluid flow with cfx
The Run made in the Previous post was or better is supposed to be a Benchmark, because the simulations for my thesis need to point to the reverse reaction, the separation of a 2 phasic system, so i wanted to at least get the generation of droplets/ mixing to work so that i have a crude understanding of the settings.

mjgraf · July 8, 2013, 09:44

you sure this was an error?
sounds like the solver met one of criteria to stop.
setting up transient there is a multitude of options for solver duration and timestep. Double check these to ensure they are set to what you require.

You might want to use Total time, of which even with adaptive timestepping should sum the total simulation time correctly.

Phantalor · July 8, 2013, 09:48

Thats why i'm posting this question ...
i used total time and he stopped at 2 seconds, with 1e5 done timesteps while he should have run until a simulation time of 5 seconds

Edit: and i already double checked i have not set a stopping criteria besides the total time of 5 seconds
Edit2: and as it was not a run end Message, so i do not believe he met a standart stopping criteria, well as said some nutjob decieded to test the electrical security of my pc without shutting it down, he crashed my hdd through unplugging the powercord and the startup HDD repair had to delete some parts so the data is nearly completely useless ... and it takes about a week to compute

ghorrocks · July 13, 2013, 08:15

That is why you ALWAYS do this for long jobs:
* Save regular full transient results files, or backup files is steady state (but they do not tend to go for weeks)
* Do a careful analysis of errors, the physics and the setup before you waste your time on a several week long run which just produces rubbish.

July 8, 2013, 06:52	Problem with maximum number of Timesteps	#1
Phantalor New Member Martin Möller Join Date: May 2013 Location: Jena, Germany Posts: 11 Rep Power: 13	Hello and good morning, I have set up a 2 Phase simulation (there is somewhere another post about the exact settingshttp://www.cfd-online.com/Forums/cfx...tml#post424961) As ghorrocks suggested in that post, i changed the timestep settings now the 2 Phase simulation works as intended. but now there is a new problem. I need to simulate not only the behavior until something happens but a timesweep of at least 5-10 seconds. The problem is because of the Timestepping settings for about 2 seconds of simulation cfx needs about 1e5 timesteps, and at this number of timesteps it exits with an error code (something like maximum number of timesteps reached) and also erases the data up to this point. (well maybe there is a backup but as my HDD crashed after the error it was erased becaus of linux HDD repair) So here is my question, is there a way to heigthen the number of timesteps while using adaptive calculations or is there a way to run the simulation step by step (also 1e5 timesteps and then the next 1e5 etc.) and is there a way to write a batch that commands ansys to do so. My Problem is that is that i do not see a way to clearly define it, because of the adaptive timestepping there is not really a way to know the time when it reaches 1e5 timesteps Ps: I'm calculating with the Academic research package (maybe its dependend on the license)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] SnappyHexMesh for internal Flow	vishwa	OpenFOAM Meshing & Mesh Conversion	24	June 27, 2016 09:54
Stable boundaries	marcoymarc	CFX	33	March 13, 2013 07:39
air bubble is disappear increasing time using vof	xujjun	CFX	9	June 9, 2009 08:59
Unaligned accesses on IA64	andre	OpenFOAM	5	June 23, 2008 11:37

July 8, 2013, 09:21		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	I do not know for sure but I do not think an academic license limits the number of time steps. I have a lot of experience in modelling MEMS microjet structures and to be honest I do not use CFX for it very much any more - it is far too slow. Have a look at Fluent. I have got Fluent running something like 20 times faster than CFX for micro drop ejection (that is free surface models with strong surface tension effects).

July 8, 2013, 09:41		#3
Phantalor New Member Martin Möller Join Date: May 2013 Location: Jena, Germany Posts: 11 Rep Power: 13	Well i would like to use fluent if it brings better results in less time, but the Pc that is for me to use and the solvers on the other machines only have cfx installed and it seems we only have a licence for fluid flow with cfx The Run made in the Previous post was or better is supposed to be a Benchmark, because the simulations for my thesis need to point to the reverse reaction, the separation of a 2 phasic system, so i wanted to at least get the generation of droplets/ mixing to work so that i have a crude understanding of the settings.

July 8, 2013, 09:44		#4
mjgraf Senior Member Join Date: Dec 2009 Posts: 131 Rep Power: 19	you sure this was an error? sounds like the solver met one of criteria to stop. setting up transient there is a multitude of options for solver duration and timestep. Double check these to ensure they are set to what you require. You might want to use Total time, of which even with adaptive timestepping should sum the total simulation time correctly.

July 8, 2013, 09:48		#5
Phantalor New Member Martin Möller Join Date: May 2013 Location: Jena, Germany Posts: 11 Rep Power: 13	Thats why i'm posting this question ... i used total time and he stopped at 2 seconds, with 1e5 done timesteps while he should have run until a simulation time of 5 seconds Edit: and i already double checked i have not set a stopping criteria besides the total time of 5 seconds Edit2: and as it was not a run end Message, so i do not believe he met a standart stopping criteria, well as said some nutjob decieded to test the electrical security of my pc without shutting it down, he crashed my hdd through unplugging the powercord and the startup HDD repair had to delete some parts so the data is nearly completely useless ... and it takes about a week to compute

July 13, 2013, 08:15		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	That is why you ALWAYS do this for long jobs: * Save regular full transient results files, or backup files is steady state (but they do not tend to go for weeks) * Do a careful analysis of errors, the physics and the setup before you waste your time on a several week long run which just produces rubbish.