Clarification on time step recommendation

newbie384 · March 15, 2014, 12:36

Hi. In CFX documentation, it is said that the appropriate time step for turbomachinery simulation is 0.1/omega~1/omega, where omega is the angular velocity of the rotating domain in radian/s. Is the recommendation make based on 1 revolution for the total time?

Also, CFX documentation also says that time step dependency should be done by using fixed but small period of simulation in physical time. Is there any guide to determine how small should the period of simulation in physical time?

Thank you in advance.

ghorrocks · March 15, 2014, 16:47

The 0.1/omega to 1.0/omega refers to simply the rotation speed in radians per second. But note this is only for steady state simulations, and is just a guide. In a steady state simulation you should use the largest physical time step size which gives you reliable convergence. There is no need to do a sensitivity analysis on time step size for a steady state simulation, you do the sensitivity analysis on the degree of convergence (usually defined as either the residual or imbalance).

For a transient simulation you should set the time step size based on a sensitivity analysis. Alternately you can use adaptive time stepping homing in on 3-5 coeff loops per iteration. I prefer the adaptive time stepping approach for most applications as a) it is on less variable to establish sensitivity on and b) The solver will quickly find its own time step size, even if your initial guess was miles off.

newbie384 · March 15, 2014, 21:48

Hi. Thank you so much for your advise. I will try the adaptive timestep. Also, is there any guide to decide the total time for transient simulation, e.g. in turbomachinery? CFX documentation only says the total time should be a small period and I do not know how small should it be.

ghorrocks · March 16, 2014, 04:40

The total time completely depends on what you are doing. You might be only interested in the starting transient, or you might need the periodic repeating flow after it settles down. Obviously the periodic flow will require longer - exactly how long depend on the flow. So put a monitor point watching a value of importance to you and run it until it reaches a repeating cycle to a tolerance you are happy with.

But note this is only of significance to transient rotor/stator models. Most turbomachinery models can be done either frozen rotor or stage and they can be done steady state - so when it is converged you have the final flow.

March 15, 2014, 12:36	Clarification on time step recommendation	#1
newbie384 Member Join Date: Mar 2013 Posts: 68 Rep Power: 13	Hi. In CFX documentation, it is said that the appropriate time step for turbomachinery simulation is 0.1/omega~1/omega, where omega is the angular velocity of the rotating domain in radian/s. Is the recommendation make based on 1 revolution for the total time? Also, CFX documentation also says that time step dependency should be done by using fixed but small period of simulation in physical time. Is there any guide to determine how small should the period of simulation in physical time? Thank you in advance.

March 15, 2014, 16:47		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	The 0.1/omega to 1.0/omega refers to simply the rotation speed in radians per second. But note this is only for steady state simulations, and is just a guide. In a steady state simulation you should use the largest physical time step size which gives you reliable convergence. There is no need to do a sensitivity analysis on time step size for a steady state simulation, you do the sensitivity analysis on the degree of convergence (usually defined as either the residual or imbalance). For a transient simulation you should set the time step size based on a sensitivity analysis. Alternately you can use adaptive time stepping homing in on 3-5 coeff loops per iteration. I prefer the adaptive time stepping approach for most applications as a) it is on less variable to establish sensitivity on and b) The solver will quickly find its own time step size, even if your initial guess was miles off. eng.abdul likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
High Courant Number @ icoFoam	Artex85	OpenFOAM Running, Solving & CFD	11	February 16, 2017 13:40
Rapidly decreasing deltaT for interDyMFoam	chrisb2244	OpenFOAM Running, Solving & CFD	3	July 1, 2014 16:40
Micro Scale Pore, icoFoam	gooya_kabir	OpenFOAM Running, Solving & CFD	2	November 2, 2013 13:58
calculation stops after few time steps	sivakumar	OpenFOAM Running, Solving & CFD	7	March 17, 2013 06:37
Full pipe 3D using icoFoam	cyberbrain	OpenFOAM	4	March 16, 2011 09:20

March 15, 2014, 21:48		#3
newbie384 Member Join Date: Mar 2013 Posts: 68 Rep Power: 13	Hi. Thank you so much for your advise. I will try the adaptive timestep. Also, is there any guide to decide the total time for transient simulation, e.g. in turbomachinery? CFX documentation only says the total time should be a small period and I do not know how small should it be.

March 16, 2014, 04:40		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,700 Rep Power: 143	The total time completely depends on what you are doing. You might be only interested in the starting transient, or you might need the periodic repeating flow after it settles down. Obviously the periodic flow will require longer - exactly how long depend on the flow. So put a monitor point watching a value of importance to you and run it until it reaches a repeating cycle to a tolerance you are happy with. But note this is only of significance to transient rotor/stator models. Most turbomachinery models can be done either frozen rotor or stage and they can be done steady state - so when it is converged you have the final flow.