Problems with Small Values in the Rigid Body Solver

dae4god · September 24, 2012, 22:41

My simulation involves rotating a rigid body at very small angular accelerations of the order of $\alpha = 4.18\times 10^{-3} rad s^{-2}$ . However, when I look at the Angular Acceleration Z in one of the CFD-Solver monitors (see screenshot), it shows $\alpha = 0 rad s^{-2}$ .

Does anyone know why this happens, or how I can rotate my rigid body at small angular accelerations?

ghorrocks · September 25, 2012, 08:11

Have you zoomed in on the graph axes? You cannot see the small number with such a wide range.

dae4god · September 25, 2012, 08:58

Great suggestion!

I just looked at the CFX-Solver monitor again and I'm not sure how to zoom in. Instead, I exported the plot data and looked at the individual numerical values. Unfortunately, the Angular Acceleration still turns out to be $\alpha=0.00E0 rad s^{-2}$ .

It still seems as if the Rigid Body Solver won't accept small values of Angular Acceleration.

ghorrocks · September 25, 2012, 19:00

Right click on the chart and set the properties to zoom in.

I would be very surprised if there is a minimum angular acceleration which it accepts. I suspect your issue is more that you are not extracting the acceleration from the solver correctly.

Can you post your CCL?

dae4god · September 25, 2012, 21:07

Ah, that's a clever zoom feature!

I've zoomed in and the Angular Acceleration still looks like it is sitting on 0 rad/s^2. I've attached another screenshot.

I wasn't sure if you wanted the actual CFX-Pre .CCL file, or just wanted to see the settings I used. I've done the latter and attached the CFX-Solver Output in the .txt file which shows the settings I used. If you would like the actual CFX-Pre .CCL file, let me know.

Thanks for all your suggestions and help!

hamed.majeed · September 25, 2012, 22:02

Hi,
give the complete output file, also attach a pic showing rigid body convergence with force, torque, and mesh motion convergence.

dae4god · September 25, 2012, 22:23

Hi Hamed,

I've attached the complete CFX-Solver Output and a screenshot of the Torque and Motion convergence (there are no forces applied on the Rigid Body so there is no Force convergence plot).

Also, I have tried running the same simulation using different Rigid Body Control settings including:

Every Timestep
Every Coefficient Loop
General Coupling Control
With and without Simo Wong Angular Momentum control

As you can see from the Output File, the screenshots in this thread are for the General Coupling Control case. Only a couple of timesteps were solved, but I can restart the simulation to test anyone's theories.

hamed.majeed · September 25, 2012, 22:43

Hi,
I had my own issues regarding rigid body convergence. I have also tried all options but rigid body does not show any convergence.
http://www.cfd-online.com/Forums/cfx...nce-issue.html

dae4god · September 25, 2012, 22:50

Hi Hamed,

Thanks for your reply and I had a look at your post. I believe you can set the convergence target for the motion, force and torque; similar to setting target residuals for mass and momentum.

Under the Rigid Body Control:

Select either Every Coefficient Loop or General Coupling Control
Check Internal Coupling Data Transfer Control
Check Mesh Motion/Force/Torque Data Transfer Control
Check Convergence Target
Set Convergence Target

Was this what you were asking in your other thread? Have you tried using these settings?

hamed.majeed · September 25, 2012, 23:09

I applied residuals for torque, force, and mesh motion. Still divergent.

hamed.majeed · September 25, 2012, 23:17

Also I tried.

hamed.majeed · September 25, 2012, 23:22

I also tried.

dae4god · September 26, 2012, 01:37

For the torque, force and motion convergence, try bumping up the maximum number of iterations to say 100. Does this change the results?

hamed.majeed · September 26, 2012, 14:38

Hi,

I have tried that also, results have improved yet same behavior.
I haven't explained my case. A rigid body falls under gravity in a column of fluid.

As I have shown that the rigid body convergence diverges. However, in ANSYS help I found this.

Force Convergence

It is defined as ratio of difference of forces for previous and present time step divided by the greater of the two forces.
Now, in my case the rigid body falls under gravity.
At each time step the force acting on the body differs as the rigid body rotates and translates. Is that the explanation of the curve behavior???

Also torque convergence is defined similar to force convergence but for torque.

dae4god · October 1, 2012, 19:57

Hi Hamed,

This is a screenshot of my Rigid Body Convergence for the simulation, run over a much longer time. You can see that the motion convergence is quite smooth, but the torque convergence is a bit more scattered.

Does this help?

hamed.majeed · October 4, 2012, 00:35

Hi,
Thanks for the reply. Is this your final curve. I mean is your solution convergent for the curves shown?
How can we say the torque of rigid body converges in this scenario!
What were the rigid body settings you used.
Thank you

Regards
Hamed

dae4god · October 4, 2012, 01:41

Hi Hamed,

The image in my last post shows the convergence levels for Motion and Torque, once all the other (mass and momentum) residuals had converged to 1E-06.

Not exactly sure if we can say that the torque and motion have converged.

You can see my Rigid Body settings from the CFX-Solver Output I posted earlier in the thread.

hamed.majeed · October 4, 2012, 02:34

Hi,
That means the solution is not acceptable. As rigid body has not converged... shouldn't it converge for solution to be acceptable?
Thanks

ghorrocks · October 4, 2012, 07:44

But how tight convergence is required? It may already be sufficiently converged.

A sensitivity study would help here, but as you do not appear to be able to converge tighter that might be difficult.

Of course I suggested ages ago you do a test with an initial condition of cross flow to see if this is a low Reynolds number affect but you appear to have ignored that suggestion.

hamed.majeed · October 5, 2012, 07:40

I will try that, thnx ghorrocks

September 25, 2012, 22:50		#9
dae4god New Member Josh Yen Join Date: Jul 2011 Posts: 14 Rep Power: 14	Hi Hamed, Thanks for your reply and I had a look at your post. I believe you can set the convergence target for the motion, force and torque; similar to setting target residuals for mass and momentum. Under the Rigid Body Control: Select either Every Coefficient Loop or General Coupling Control Check Internal Coupling Data Transfer Control Check Mesh Motion/Force/Torque Data Transfer Control Check Convergence Target Set Convergence Target Was this what you were asking in your other thread? Have you tried using these settings?

October 4, 2012, 07:44		#19
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	But how tight convergence is required? It may already be sufficiently converged. A sensitivity study would help here, but as you do not appear to be able to converge tighter that might be difficult. Of course I suggested ages ago you do a test with an initial condition of cross flow to see if this is a low Reynolds number affect but you appear to have ignored that suggestion. hamed.majeed likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Difficulty in calculating angular velocity of Savonius turbine simulation	alfaruk	CFX	14	March 17, 2017 06:08
Rigid body - pivoting about a fixed point	Catenary	CFX	0	April 26, 2012 02:07
Numerical errors in nested domain with pre-calculated boundary values	Arnoldinho	OpenFOAM Running, Solving & CFD	3	April 4, 2012 10:31
Problems with SUPG body force term	FEM question	Main CFD Forum	0	January 21, 2006 17:51
CFX 5.5	Roued	CFX	1	October 2, 2001 16:49

September 25, 2012, 08:11		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Have you zoomed in on the graph axes? You cannot see the small number with such a wide range.

September 25, 2012, 08:58		#3
dae4god New Member Josh Yen Join Date: Jul 2011 Posts: 14 Rep Power: 14	Great suggestion! I just looked at the CFX-Solver monitor again and I'm not sure how to zoom in. Instead, I exported the plot data and looked at the individual numerical values. Unfortunately, the Angular Acceleration still turns out to be $\alpha=0.00E0 rad s^{-2}$ . It still seems as if the Rigid Body Solver won't accept small values of Angular Acceleration.

September 25, 2012, 19:00		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Right click on the chart and set the properties to zoom in. I would be very surprised if there is a minimum angular acceleration which it accepts. I suspect your issue is more that you are not extracting the acceleration from the solver correctly. Can you post your CCL?

September 25, 2012, 22:02		#6
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Hi, give the complete output file, also attach a pic showing rigid body convergence with force, torque, and mesh motion convergence.

September 25, 2012, 22:43		#8
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Hi, I had my own issues regarding rigid body convergence. I have also tried all options but rigid body does not show any convergence. http://www.cfd-online.com/Forums/cfx...nce-issue.html

September 25, 2012, 23:09		#10
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	I applied residuals for torque, force, and mesh motion. Still divergent.

September 25, 2012, 23:17		#11
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Also I tried.

September 25, 2012, 23:22		#12
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	I also tried.

September 26, 2012, 01:37		#13
dae4god New Member Josh Yen Join Date: Jul 2011 Posts: 14 Rep Power: 14	For the torque, force and motion convergence, try bumping up the maximum number of iterations to say 100. Does this change the results?

September 26, 2012, 14:38		#14
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Hi, I have tried that also, results have improved yet same behavior. I haven't explained my case. A rigid body falls under gravity in a column of fluid. As I have shown that the rigid body convergence diverges. However, in ANSYS help I found this. Force Convergence It is defined as ratio of difference of forces for previous and present time step divided by the greater of the two forces. Now, in my case the rigid body falls under gravity. At each time step the force acting on the body differs as the rigid body rotates and translates. Is that the explanation of the curve behavior??? Also torque convergence is defined similar to force convergence but for torque.

October 4, 2012, 00:35		#16
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Hi, Thanks for the reply. Is this your final curve. I mean is your solution convergent for the curves shown? How can we say the torque of rigid body converges in this scenario! What were the rigid body settings you used. Thank you Regards Hamed

October 4, 2012, 01:41		#17
dae4god New Member Josh Yen Join Date: Jul 2011 Posts: 14 Rep Power: 14	Hi Hamed, The image in my last post shows the convergence levels for Motion and Torque, once all the other (mass and momentum) residuals had converged to 1E-06. Not exactly sure if we can say that the torque and motion have converged. You can see my Rigid Body settings from the CFX-Solver Output I posted earlier in the thread.

October 4, 2012, 02:34		#18
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	Hi, That means the solution is not acceptable. As rigid body has not converged... shouldn't it converge for solution to be acceptable? Thanks

October 5, 2012, 07:40		#20
hamed.majeed Senior Member Hamed Abdul Majeed Join Date: May 2012 Location: New Orleans, LA, US Posts: 147 Rep Power: 13	I will try that, thnx ghorrocks