Timescale

MissCFD · October 21, 2015, 17:03

Can someone explain me why CFX says that if the pseudo time scale is too small, the solution may not be fully accurate ?

highorder_cfd · October 21, 2015, 18:06

Where and when does CFX say this? You should post more details in order to get a reply.

For example, if you are solving a steady state problem and the time step is too small, it is possible you will solve unsteady behaviours affecting your
solution and thus your final steady state solution will not be properly converged.

ghorrocks · October 21, 2015, 19:00

In addition to highorder's comment, there is also:
* round off errors can affect small time step sizes
* the very slow convergence associated with small time steps can fool some convergence criteria into thinking the solution is converged prematurely. This depends on how the convergence criteria is defined.

MissCFD · October 21, 2015, 23:41

Thank you very much for your answer.

When you say "round off errors", it's because there are more timescale so we accumulate many more errors ?

I don't see/understand very well your second point. Can you give me a concrete example to illustrate it ?

Quote:

Originally Posted by ghorrocks

In addition to highorder's comment, there is also:
* round off errors can affect small time step sizes
* the very slow convergence associated with small time steps can fool some convergence criteria into thinking the solution is converged prematurely. This depends on how the convergence criteria is defined.

ghorrocks · October 22, 2015, 05:34

Round off errors - the main thing is that as the divisor get smaller the accuracy of the calculation decreases. This applies more to the mesh and physics than the time step, but can apply to the time step in some circumstances.

Convergence can be defined by when an output parameter changes by less than x% over the last 100 iterations. If you use a really small time step you can artificially slow this down and trigger your convergence criteria prematurely.

October 21, 2015, 17:03	Timescale	#1
MissCFD Senior Member Join Date: Aug 2014 Posts: 160 Rep Power: 11	Can someone explain me why CFX says that if the pseudo time scale is too small, the solution may not be fully accurate ?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
When to use local timescale or physical timescale	xyq102296	CFX	23	June 25, 2014 07:55
CFX Timescale: Difference between Physical,Local facotr and Auto Timescale	Mfaizan	CFX	4	April 7, 2014 02:03
Timescale Update Frequency	100tinela	CFX	5	January 25, 2013 07:02
Determination of physical timescale	Chander	CFX	2	October 19, 2011 19:47
Solid Timescale Control	Parthipan	CFX	2	August 24, 2007 10:07

October 21, 2015, 18:06		#2
highorder_cfd Member Join Date: Jan 2015 Posts: 63 Rep Power: 11	Where and when does CFX say this? You should post more details in order to get a reply. For example, if you are solving a steady state problem and the time step is too small, it is possible you will solve unsteady behaviours affecting your solution and thus your final steady state solution will not be properly converged.

October 21, 2015, 19:00		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	In addition to highorder's comment, there is also: * round off errors can affect small time step sizes * the very slow convergence associated with small time steps can fool some convergence criteria into thinking the solution is converged prematurely. This depends on how the convergence criteria is defined.

October 22, 2015, 05:34		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Round off errors - the main thing is that as the divisor get smaller the accuracy of the calculation decreases. This applies more to the mesh and physics than the time step, but can apply to the time step in some circumstances. Convergence can be defined by when an output parameter changes by less than x% over the last 100 iterations. If you use a really small time step you can artificially slow this down and trigger your convergence criteria prematurely.