Transient timestep question

AZ1621 · July 13, 2015, 17:53

Hello,

I am running a transient turbulent flow simulation of a bluff body in a wind tunnel. I am trying to validate experimental data with my results. To get my residuals down to the appropriate level I have to use an extremely small time step (1 e-5 seconds). I am trying to model and time average the flow properties surrounding the vortex shedding occurring in my simulation but using the time-step of 1e-5 it would take a huge amount of time-steps to gather data for ten or twelve shedding cycles due to the startup time for the shedding to begin occurring from the moment the simulation begins. I thought to use a larger time step to achieve a flow time where I know shedding is occurring but I don't converge above the 1e-5 second timestep setting. Can I still use larger timesteps to advance? Or is there a better way to go about this?

Thanks for the help!

LuckyTran · July 15, 2015, 11:45

You really only need a small enough time-step to temporally resolve the time-scale you are interested in (the shedding frequency I guess). This can be done with as few as 10 time-steps per event. You should compare the 1e-05 to the time-scale of the shedding frequency.

But of course it is important to make sure that the time-steps are converged.

AZ1621 · July 15, 2015, 15:00

Thanks for the reply,
Last night I simulated 500 iterations at 0.01(s) timestep. I observed the vortex shedding had a period of about 10 timesteps or 0.1(s) so the shedding frequency is about 10Hz. Now all the residuals on those iterations ran for the full 20 inner iterations and only got down to about 1e-2. So running a single shedding period at 1e-5(s) timesteps would take about 10,000 iterations. I want to time average the shedding so I will need 8-10 cycles. This ups my iterations to about 80,000-90,000. This would take an enormous amount of time to simulate even if the iterations converge every 8 inner iterations. Could be an issue to do because my licence resets every 4 hours. Any thoughts on how to reduce increase the timestep while still retaining convergence?

Only super concerned with using the very small timestep because at a timestep of 0.01(s) i was seeing about a 18.75% difference in the mean values when compared with experimental data.

July 13, 2015, 17:53	Transient timestep question	#1
AZ1621 New Member Join Date: Jun 2015 Posts: 7 Rep Power: 10	Hello, I am running a transient turbulent flow simulation of a bluff body in a wind tunnel. I am trying to validate experimental data with my results. To get my residuals down to the appropriate level I have to use an extremely small time step (1 e-5 seconds). I am trying to model and time average the flow properties surrounding the vortex shedding occurring in my simulation but using the time-step of 1e-5 it would take a huge amount of time-steps to gather data for ten or twelve shedding cycles due to the startup time for the shedding to begin occurring from the moment the simulation begins. I thought to use a larger time step to achieve a flow time where I know shedding is occurring but I don't converge above the 1e-5 second timestep setting. Can I still use larger timesteps to advance? Or is there a better way to go about this? Thanks for the help!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Transient or Transient Blade Row?	liucheng8602	CFX	2	June 19, 2014 07:03
transient timestep	tony	CFX	1	March 12, 2008 16:39
Time value of timestep in Transient simulations	siv	CFX	3	December 8, 2005 11:15
Timestep for transient simulation	houman	CFX	2	April 28, 2005 03:44
timestep duration-a question	Shankar	FLUENT	2	August 15, 2002 12:12

July 15, 2015, 11:45		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 65	You really only need a small enough time-step to temporally resolve the time-scale you are interested in (the shedding frequency I guess). This can be done with as few as 10 time-steps per event. You should compare the 1e-05 to the time-scale of the shedding frequency. But of course it is important to make sure that the time-steps are converged.

July 15, 2015, 15:00		#3
AZ1621 New Member Join Date: Jun 2015 Posts: 7 Rep Power: 10	Thanks for the reply, Last night I simulated 500 iterations at 0.01(s) timestep. I observed the vortex shedding had a period of about 10 timesteps or 0.1(s) so the shedding frequency is about 10Hz. Now all the residuals on those iterations ran for the full 20 inner iterations and only got down to about 1e-2. So running a single shedding period at 1e-5(s) timesteps would take about 10,000 iterations. I want to time average the shedding so I will need 8-10 cycles. This ups my iterations to about 80,000-90,000. This would take an enormous amount of time to simulate even if the iterations converge every 8 inner iterations. Could be an issue to do because my licence resets every 4 hours. Any thoughts on how to reduce increase the timestep while still retaining convergence? Only super concerned with using the very small timestep because at a timestep of 0.01(s) i was seeing about a 18.75% difference in the mean values when compared with experimental data.