Force on body oscillation

 Register Blogs Members List Search Today's Posts Mark Forums Read

January 9, 2020, 03:55
Force on body oscillation
#1
New Member

Join Date: Apr 2014
Posts: 11
Rep Power: 8
Dear all;
I am simulating a problem in which the force_y on body is calculated every time step. Also because of mesh motion, I am using the remeshing.
Everything is fine until remeshing time step. After the remeshing, the force_y on body (and any forces on body) is oscillate.
After some steps, the oscillations decrease until the next remeshing begins. The force's diagram is attached. The remeshing steps are marked with a arrow.
Is the problem of interpolation after remeshing? If yes, how can I improve the accuracy of the interpolation?
I check the geometry and mesh, the remeshing is done with great accuracy.
Attached Images
 2020-01-09_12-20-13.png (54.3 KB, 14 views)

 January 9, 2020, 05:52 #2 Senior Member   Gert-Jan Join Date: Oct 2012 Posts: 1,155 Rep Power: 17 Is this a transient calculation?

January 9, 2020, 06:04
#3
New Member

Join Date: Apr 2014
Posts: 11
Rep Power: 8
Quote:
 Originally Posted by Gert-Jan Is this a transient calculation?

Hi Dear Gert-Jan,

 January 9, 2020, 07:01 #4 Senior Member   Gert-Jan Join Date: Oct 2012 Posts: 1,155 Rep Power: 17 Yeah, you mentioned Mesh motion, sorry......... But what is on the x-axis? Do we see 50 timesteps (coefficient loops not visible) where at timesteps 14, 26 and 40 remeshing is performed? This would mean that in the intermediate timesteps remeshing is not executed and you have only deforming mesh? Then, the options to test are: reduce timestep size, remesh more often, increase the number of coefficient loops within a timestep, tighten the convergence critera (default 1e-4 RMS). Probably there are more, but I have little experience in transient remeshing. m_mehrdad likes this.

January 9, 2020, 08:13
#5
New Member

Join Date: Apr 2014
Posts: 11
Rep Power: 8
Quote:
 Originally Posted by Gert-Jan Yeah, you mentioned Mesh motion, sorry......... But what is on the x-axis? Do we see 50 timesteps (coefficient loops not visible) where at timesteps 14, 26 and 40 remeshing is performed? This would mean that in the intermediate timesteps remeshing is not executed and you have only deforming mesh? Then, the options to test are: reduce timestep size, remesh more often, increase the number of coefficient loops within a timestep, tighten the convergence critera (default 1e-4 RMS). Probably there are more, but I have little experience in transient remeshing.

The x-axis is accumulated time step. Yes, at timesteps 13, 26 and 39 remeshing is performed. I define a RemeshCriterion for Interrupt condition as follow:
minVal(Orthogonality Angle Minimum )@Fluid<10[deg]

The diagram of Mesh_Quality and RemeshCriterion is attached. In remeshing problems, simulation continue until Interrupt condition (Remeshcriterion). In the intermediate timesteps remeshing is not executed and the simulation only deform the mesh. In Timestep of Interrupting the run is stopped and the geometry and mesh are renew. The new mesh file is send to CFX-Pre and the last solution in interpolated and the run continues.

I test with more fine mesh, reduce timestep size, And also i fixed a part of the body, However the force on this part was also oscillate with remeshing.

Can I improve ,the interpolation of the last solution, when remeshing?
Attached Images
 2020-01-09_16-32-51.png (25.9 KB, 5 views)

 January 9, 2020, 09:49 Improve Interpolation after remeshing. #6 New Member   Mehrdad Join Date: Apr 2014 Posts: 11 Rep Power: 8 Please help me for Improve Interpolation after remeshing.

 January 10, 2020, 03:10 #7 New Member   Mehrdad Join Date: Apr 2014 Posts: 11 Rep Power: 8 Can anyone help me?

 January 19, 2020, 03:35 #8 Super Moderator   Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 16,315 Rep Power: 125 We are not going to be able to help you as the problem is going to be some detail of your setup. We cannot analyse your setup in detail, only you can do this. But before proceeding you should do a sensitivity analysis on mesh size, timestep size and convergence criteria. There is no point in looking at the results of a simulation where the results are horribly inaccurate as the basics are wrong. And once you have done this you will probably need to do a second sensitivity analysis looking at remeshing frequency and settings. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.

January 19, 2020, 04:18
#9
New Member

Join Date: Apr 2014
Posts: 11
Rep Power: 8
Quote:
 Originally Posted by ghorrocks We are not going to be able to help you as the problem is going to be some detail of your setup. We cannot analyse your setup in detail, only you can do this. But before proceeding you should do a sensitivity analysis on mesh size, timestep size and convergence criteria. There is no point in looking at the results of a simulation where the results are horribly inaccurate as the basics are wrong. And once you have done this you will probably need to do a second sensitivity analysis looking at remeshing frequency and settings.

Dear Glenn Horrocks
Thank you very much. I do a sensitivity analysis on mesh size, timestep size and convergence criteria. Also, I do a sensitivity analysis on remeshing frequency and settings.Every things is ok but after remeshing and interpolation this problem is occured.

When interpolating, the data of app. 98% of element are mapped and only the dataof app. 2% of element are interpolated.

My question is, Is this interpolation the reason for some parameters to fluctuate when remeshing?

Have you ever encountered remeshing simulations?

If yes, Is it possible to check the forces on the body during the remeshing?

After much analysis I have come to the conclusion that these fluctuations at remeshing time are quite natural and there is no way to eliminate, smooth and reduce them. Is this theory correct?

 January 19, 2020, 04:32 #10 Super Moderator   Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 16,315 Rep Power: 125 I have never user dynamic remeshing so cannot comment. But thinking about how remeshing works, I suspect it is not a conservative interpolation. In other words, there is nothing to ensure that quantities are conserved at any length scale before and after meshing. This means you would expect variables to jump on remeshing. But it also means that the size of the jump should decrease with mesh refinement. So this might mean you require a finer mesh than you normally do to get the remeshing jumps small enough to be acceptable. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.