Rigid body + remeshing error in redefining the rigid body updated-initial conditions

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

April 18, 2012, 09:46
Rigid body + remeshing error in redefining the rigid body updated-initial conditions
#1
Member

Patrick Wollny
Join Date: Apr 2010
Posts: 58
Rep Power: 12
Hello,

I'm trying to simulate a ball inside a diffuser being lifted by the air blowing upwards from the bottom opening of the diffusor. Gravitation is taken into account in downward direction. The mesh is allowed to deform with respect to the position of the ball, at a point when the deformation is to an extent (minimum othogonal angle < 15 deg), then the domain is allowed to be remeshed by a ICEM replay file. The remeshing works fine, but then the newly meshed domain failes to read the new updated initial boundar conditions which where supposed to be the latest values of the last grid (the rigid body velocity and acceleration). The acceleartion and velocities seem to be redefined automatically from some mystry values (not the expected ones)

However the position of the ball seems to be read correctly, excpet for the velocities and accelaration of the ball.

Does someone know the solution to read in the proper values for acceleration and velocity and avoid these rather unknown values instead?

PS: the posted grafs can be seen to notice the above mentioned problem.
Attached Images
 RigidBodyRemesh.jpg (39.1 KB, 82 views) RigidBodyVelo.jpg (31.0 KB, 54 views)

 April 18, 2012, 10:43 #2 Senior Member   Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 17 How are you moving the ball? Are you using the 6DOF solver or moving it through CEL based on exterior forces?

 April 18, 2012, 10:57 #3 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 I move the ball by the 6 DOF-Solver, but I only solve translational motions (CFX standard newmark Newmark-Method).

 April 18, 2012, 11:00 #4 Senior Member   Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 17 Ok. I was going to suggest if it was by CEL, was that you were not carrying the momentum through the remesh, but the 6DOF will do that for you. How are you specifying your BCs? Are they constant or time dependant?

 April 18, 2012, 11:04 #5 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 The BCs are all constand. Only the ball should move in space.

 April 18, 2012, 11:27 #6 Senior Member   Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 17 I assume you are using rbstate to pull the acceleration and velocity out. Right before remesh it looks like your velocity wants to steady out and acceleratoin might want to tend to loiter around 0. I suggest reducing the remesh state to mesh orthogonality angle <5 (or perhap just turn off the remesh state and let it run till you get to negative volumes). Does the velocity really steady out? Next I would at least double your mesh density and rerun the same condition. Does V and a profile look the same? If not, you haven't achieved mesh independence, and remeshing can really show this. In addition, are you letting the ball baffle around in the diffusor or should this be basically linear motion (one direction)? If it is only one direction, I would adjust your mesh deformation scheme to a smarter method that would allow you to move the ball much greater distances without remesh (you might be able to do the entire motion without remesh).

 April 18, 2012, 12:06 #7 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 Thank you for your answer, I only let the ball move up and down to reach a equilibrium point between the gravitation and the fluid impulse. Yes I could model the mesh motion on a smarter way, but this is just a test, later I would like to simulate the transport of a rigid body by the air. This will be a much complexer motion. I will simulate a case without remeshing and try to increase the mesh density. My mesh is not very fine, but is it possible, that a coarse grid leads to such an error?

 April 18, 2012, 12:12 #8 Senior Member   Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 17 Very possible. Course mesh works well to make sure the motion is going ok, less well to keep things like what you are seeing in check.

April 18, 2012, 13:57
#9
Member

Patrick Wollny
Join Date: Apr 2010
Posts: 58
Rep Power: 12
Hi Singer,

here are the results of the simulation without remeshing. As I expected the acceleration increases strongly, because the velocity at the inlet is more than twice times higher than the equilibrium velocity.
Attached Images
 WithoutRemeshAcc.jpg (37.7 KB, 38 views) withoutRemeshVelo.jpg (35.5 KB, 31 views)

April 19, 2012, 09:31
#10
Member

Patrick Wollny
Join Date: Apr 2010
Posts: 58
Rep Power: 12
I did the simulation again with a refined grid. The refined mesh has nearly 500000 cells in a diffusor with a length of 0.15 m and a 0.04 - 0.06m diameter. The inlet velocity is about 10 m/s. I think the grid is fine enough.

I also did a simulation with two grids:
one was a sphere which enclosed the ball. This grid moved with the Ball and was not remeshed. A second grid enclosed the spherical grid (interploation via GGI) and presented the diffusor. This grid has been remeshed from time to time.

The simulations show the same unsteadiness and the same drop of the acceleration of the ridig body after the remeshing. The remeshing itself seems not to be the problem, because the spherical mesh around the ball has not been remeshend and even then the rigid body acceleartion droped after remeshing. The deviating result in the simulation with two grids is probably because I have chosen a too small area for the spherical grid around the ball.
Attached Images
 refineAcc.jpg (40.6 KB, 23 views) 2MeshesAcc.jpg (37.0 KB, 23 views)

April 19, 2012, 10:08
#11
Super Moderator

Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 16,539
Rep Power: 128
Quote:
 I think the grid is fine enough.
I cannot count the amount of times I have heard that quoted on the forum I hope you did not simply look at the mesh and think the elements look teeny-tiny so that should be fine.

Do a mesh sensitivity check, this is the way to be sure your mesh is adequate.

April 19, 2012, 11:06
#12
Member

Patrick Wollny
Join Date: Apr 2010
Posts: 58
Rep Power: 12
Quote:
 Originally Posted by ghorrocks I cannot count the amount of times I have heard that quoted on the forum I hope you did not simply look at the mesh and think the elements look teeny-tiny so that should be fine. Do a mesh sensitivity check, this is the way to be sure your mesh is adequate.

Of course you are right, but now I just want to get basic rigid body physics redefined on the new mesh. The simulation time, for example, restarts after each remeshing from 0, if you set the initial time to 0 and not to automatic. The refinement I have done delivers no better redefinement after the remeshing, so I think that the mesh density is not the problem.

 April 19, 2012, 19:30 #13 Super Moderator   Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 16,539 Rep Power: 128 OK, that sounds like you have some evidence to show your mesh is adequate. But a final comment - how different were the meshes you compared? I halve the element edge length, which results in a mesh approximately 8 times more elements than before if applied globally. If you only make a small change in mesh size it can fool you into thinking you have converged.

 April 21, 2012, 08:20 #14 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 Hi ghorrocks, I halve the element edge length too. Now I have the plan to input the acceleration and velocity manual via junction box. I only have to know the directory where the rigid body stats are saved in the ram. Does someone know this directory or has someone a list of all CFX directories in the ram for using the mms (memory management system) ?

April 21, 2012, 08:30
#15
Super Moderator

Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 16,539
Rep Power: 128
Quote:
 I halve the element edge length too.
Ok good. Now I am convinced!

 April 25, 2012, 17:37 #16 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 Is it possible to enable the rigid body mode for a boundary and disable the rigid body solver? I would like to solve the equation of motion with my own code, but that would be easier with the rigid body data aviable in the solver (mass of inertia, torque, ...).

 April 25, 2012, 18:02 #17 Senior Member   Edmund Singer P.E. Join Date: Aug 2010 Location: Minneapolis, MN Posts: 511 Rep Power: 17 I think what you are discribing can be done using mesh motion only (forget the 6dof solver). I assume you have a purely translational movement. One trick, in order to get large mesh deformations that don't collaspe, is to utilize a subdomain around your body of interest and move the mesh of the subdomain based on translational motion that you can write the equations of motion in CEL using the force on the walls of interest. The mesh around the subdomain can be linearly ""squashed/expaneded" based on its distance from the subdomain. There should be many threads on this, so do a search on that.

 April 27, 2012, 11:58 #18 Member   Patrick Wollny Join Date: Apr 2010 Posts: 58 Rep Power: 12 The translational motion is just the beginning, the rotation will follow. Later I'll move a bigger "particle" through a channel via air flow (with 6dof). Now I try to simulate the ball in the diffusor with the immersed solid, rigid body solver in cfx. For a good result it is importend to set the "Momentum Source Scaling Factor" between 50 and 100. In this case I have to set the expert parameter "smooth inside ims = t" ( have a look at this thread: http://www.cfd-online.com/Forums/cfx/69285-question-about-immersed-solid-cfx-12-0-a.html ) . The problem is, that the cfx solver gives me an error. The setting seems to be onknown for it. Here is the error message: +--------------------------------------------------------------------+ | ERROR #001100000 has occurred in subroutine REPORT_OBSOLETE_PRM. | | Message: | | The following unused Expert Solver Parameter was found: | | | | SMOOTH INSIDE IMS | | | | The parameter may be incorrectly spelled. | +--------------------------------------------------------------------+ I have inserted the parameter via command editor in the solver tree: FLOW: Flow Analysis 1 &replace EXPERT PARAMETERS: smooth inside ims = t END END Does the setting "smooth inside ims = t" not exist anymore? I get an error with a high Momentum Source Scaling Factor + Boundary model = Modified Forcing ( Boundary Face Extrusion).

 May 25, 2012, 07:10 #19 New Member   belgacem Join Date: Jan 2012 Posts: 22 Rep Power: 11 Hi Friends I am also studying immersed boundary method and and try to simulate a block falling in the water. I am using "immersed solid" then rigid body 6DOF and I let it fall freely but it can't stoped in the bottom where the velocity must be zero. I give a density to the block and i let it fall freely under gravity. Noted that the rigid body is defined as an immersed solid. i have specified a stationary coordinate frame that has its origin at the center of mass of the physical rigid body. Another fixed coordinate frame was specified related to the water at rest. What can I do to stopped the rigid body in the bottom where the potentiel energy must be zero? thank you!

August 22, 2012, 12:04
#20
New Member

fujiaquan
Join Date: Aug 2012
Posts: 3
Rep Power: 10
Quote:
 Originally Posted by Pat84 The translational motion is just the beginning, the rotation will follow. Later I'll move a bigger "particle" through a channel via air flow (with 6dof). Now I try to simulate the ball in the diffusor with the immersed solid, rigid body solver in cfx. For a good result it is importend to set the "Momentum Source Scaling Factor" between 50 and 100. In this case I have to set the expert parameter "smooth inside ims = t" ( have a look at this thread: http://www.cfd-online.com/Forums/cfx/69285-question-about-immersed-solid-cfx-12-0-a.html ) . The problem is, that the cfx solver gives me an error. The setting seems to be onknown for it. Here is the error message: +--------------------------------------------------------------------+ | ERROR #001100000 has occurred in subroutine REPORT_OBSOLETE_PRM. | | Message: | | The following unused Expert Solver Parameter was found: | | | | SMOOTH INSIDE IMS | | | | The parameter may be incorrectly spelled. | +--------------------------------------------------------------------+ I have inserted the parameter via command editor in the solver tree: FLOW: Flow Analysis 1 &replace EXPERT PARAMETERS: smooth inside ims = t END END Does the setting "smooth inside ims = t" not exist anymore? I get an error with a high Momentum Source Scaling Factor + Boundary model = Modified Forcing ( Boundary Face Extrusion).
Same question~