[udcbr]

Hello,
I created a simulation in which is Energy applied at different positions over time, using an UDF.
After several energy applications (ca. 10-100), there is always a long break of no energy application (ca. 1000 normal energy applications long).

So I created also an UDF to define DeltaT, which is using less and longer timesteps during the breaks (for example 20 instead of 1000 timesteps) to decrease computing time.
This is running fine on my local windows machine.



However running the simulation on a Linux Cluster gives me wrong results, (temperatures lower than initial conditions, higher error, ...)

I have tried to lower the time step size or using a relaxion factor, but there are no big differences.
At some point the error "temperature limited to 1K" is happening and even before the result files of the temperature export seem to have something like waves.
Setting the temperature limit to a higher level was also not helping



My commandline commands used on the linux cluster are the following and already incorporating (rpsetvar 'temperature/secondary-gradient? #f)

file/rcd input.cas
(rpsetvar 'temperature/secondary-gradient? #f)
/solve/dual-time-iterate
1000
20
q
file/wd output.dat
exit
y



If I am using a fixed timestep size (defined in fluent or the UDF) gives me however correct results. But this is not a solution, because the computation time is too long in this case.


What can I do to obtain the same results on the Linux cluster?

[kaeran]

Hi udcbr,

I am facing the same problem but with transient flow in Fluent.
I have also posted a similar problem link in Fluent community.
Have u got rid of this problem???If so, please share the information..

[udcbr]

Originally Posted by kaeran

Hi udcbr,
I am facing the same problem but with transient flow in Fluent.
I have also posted a similar problem link in Fluent community.
Have u got rid of this problem???If so, please share the information..

Finally, I got rid of my Problem in my transient simulation by simply using NITA (Non-Iterative Time Advancement).


kaeran, you might have also a look at http://www.cfd-online.com/Forums/flu...nvergence.html