overflow and negative volume error in mesh deformation

sakalido · November 16, 2012, 18:01

Hi all,

I am working on a mesh deformation problem. It's a pressure based valve. Cold and hot water are entered from separate inlets and the piston inside the valve moves based on the cold and hot water pressure balance.

I am constantly getting either "negative volume" error or "overflow".
I know this a common error in problems involved with mesh deformation.
I have read the forum and have tried many things but nothing makes any difference.

1. My mesh is pretty fine.
2. I have decreased time step.
3. Tried 1/wall distance and 1/finite volumes and also increase near small volumes and increase near boundaries for mesh stiffness.
4. used under relaxation
5. have tried k-e and SST turbulence models

my inlet and outlet B.C's are all opening and relative pressure. For now I'm defining a sinusoidal time dependent piston movement.

I don't know whether the errors are related to the physics of the problem or mesh deformation (i.e. unspecified vs. specific displacements)?
Do you have any suggestions as to what I should check again?

Thank you,
Sara

sakalido · November 16, 2012, 18:23

and I have tried "inlet" boundary condition as well with specifying mass flow.

Again no difference.

cdegroot · November 17, 2012, 15:25

I think the problem is with the mesh deformation, not the physics. Do you expect large deformations? Post a picture of you geometry. Sometimes you can use a GGI interface which allows the meshes to not line up in order to avoid mesh tangling, but it depends on the geometry. I am thinking maybe you can create a region around your valve that simply moves as a rigid body and connects to the rest of the geometry via GGI. I'd have to see the geometry though to know if this makes sense.

ghorrocks · November 17, 2012, 20:00

Chris is right, this error has nothing to do with the physics. The mesh motion has been sufficient to turn elements inside out.

I recommend you do a run were you save lots of results files (every time step if possible) and include the mesh. If the motion is not coupled to the flow then turn the flow solvers off and just run mesh motion. Then you can see the mesh as it deforms and you should be able to identify what region is causing the problem. Once you know where the problem is you can think of options to fix it.

sakalido · November 27, 2012, 10:39

Chris and Glen,

I attached a picture of my valve geometry.
I highlighted the moving walls (piston and the washer in between two sides of the piston which I have specified mes deformation) and also the affected wall (unspecified mesh deformation). The rest of the geometry is stationary. Except for symmetry wall which all have unspecified mesh deformation.

The GGI interface that you mentioned is the default mesh connection method in my CFX simulation.

my time step is 10^-4 and my simulation stops in the first few iteration either by overflow error or negative volume error.

As I mentioned in my first thread, I have tried many things suggested in the forum but so far none of them has made any difference.

Do you have any suggestions?

cdegroot · November 27, 2012, 11:11

I just want to clarify that the moving part is separate from the stationary part and is connected using GGI. Because if they were in the same part the mesh would definitely get all tangled up. I think Glenn's idea is best: save results every timestep and see what happens to the mesh. Feel free to post the result here. Just to one iteration per timestep since all you would be trying to see is what is going wrong with the mesh.

sakalido · November 27, 2012, 12:04

Chris,

The right and left pistons are a separate part and the main domain inside the valve is a separate part as well. The washer in between the right and left pistons is a subdomain inside the main domain.

I will try to run the simulations and stop it before it overflows to check the mesh deformation shape.

Sara

cdegroot · November 27, 2012, 12:22

Instead of trying to stop it you can just have it write an output file every iteration since you said it doesn't do very many. In Pre, go to Output Control > Backup and set output frequency to every iteration.

sakalido · November 27, 2012, 13:58

I have four iterations now and I looked at my mesh and tried to make a video of the mesh deformation. I don't see any strange deformation since the movement is very small. Unfortunately the saved video is not helpful to post here bc it doesn't show the mesh, it just shows a blank solid surface!

I also looked at the mesh deformation contours, the maximum deformation is 2^-6.

How can I check whether I have tangled mesh?

cdegroot · November 27, 2012, 14:19

I think you should be able to visually see if there are problems. In CFD Post experiment with showing the surface mesh of various parts and look for anything out of the ordinary. There are also various variables you can plot related to the mesh (aspect ratio, face angle, etc.). Sounds like your maximum deformation is not that big so far, so maybe there is some inconsistency in the way the mesh motion is specified?? Maybe you should post more details about how you have set that up.

sakalido · November 27, 2012, 15:16

Chris,

I agree with you. I think the problem is with the inconsistencies of the movements I define. I attached a picture. Here is more detail:

1. I have marked the walls with specified mesh deformation in my simulation.
They include the left and right sides of the washer, the left and right side walls of the piston in contact with washer. and the default fluid fluid interface side 1 and 2 (which is the right and left piston interface with the valve main domain). The movement is defined as: 3[mm]*sin(2*pi*time)

2. That part of default domain (specified in the attached picture) which is in direct contact with washer is also moving with unspecified mesh deformation.

3. I have two domains: piston and the main valve domain. washer is a subdomain in the main valve domain. subdomain also has the same specified movement.

4. all symmetry boundary conditions have unspecified mesh deformation.

5. The rest of the main valve domain is stationary. Should I define unspecified mesh deformation for this as well?

Thank you very much for your time and help,
Sara

cdegroot · November 27, 2012, 16:16

That image helps explains things for sure. I have questions still about the way the motion is specified on the regions with specified motion. Is the displacement set to be relative to the initial mesh? I think that is what you intend. Are you using the specified displacement option on those faces? Double check that both sides of the interface are specified in a consistent way so that they move in the same direction, not opposite each other. I also wonder if the "affected wall" should be stationary. Seems to me like a rigid part.

November 16, 2012, 18:01	overflow and negative volume error in mesh deformation	#1
sakalido Member Sara S. Join Date: May 2010 Location: NJ, USA Posts: 41 Rep Power: 16	Hi all, I am working on a mesh deformation problem. It's a pressure based valve. Cold and hot water are entered from separate inlets and the piston inside the valve moves based on the cold and hot water pressure balance. I am constantly getting either "negative volume" error or "overflow". I know this a common error in problems involved with mesh deformation. I have read the forum and have tried many things but nothing makes any difference. 1. My mesh is pretty fine. 2. I have decreased time step. 3. Tried 1/wall distance and 1/finite volumes and also increase near small volumes and increase near boundaries for mesh stiffness. 4. used under relaxation 5. have tried k-e and SST turbulence models my inlet and outlet B.C's are all opening and relative pressure. For now I'm defining a sinusoidal time dependent piston movement. I don't know whether the errors are related to the physics of the problem or mesh deformation (i.e. unspecified vs. specific displacements)? Do you have any suggestions as to what I should check again? Thank you, Sara

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Mesh orientation and negative volumes	Nick R	CFX	3	January 10, 2011 17:34

November 16, 2012, 18:23		#2
sakalido Member Sara S. Join Date: May 2010 Location: NJ, USA Posts: 41 Rep Power: 16	and I have tried "inlet" boundary condition as well with specifying mass flow. Again no difference.

November 17, 2012, 15:25		#3
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	I think the problem is with the mesh deformation, not the physics. Do you expect large deformations? Post a picture of you geometry. Sometimes you can use a GGI interface which allows the meshes to not line up in order to avoid mesh tangling, but it depends on the geometry. I am thinking maybe you can create a region around your valve that simply moves as a rigid body and connects to the rest of the geometry via GGI. I'd have to see the geometry though to know if this makes sense.

November 17, 2012, 20:00		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,854 Rep Power: 144	Chris is right, this error has nothing to do with the physics. The mesh motion has been sufficient to turn elements inside out. I recommend you do a run were you save lots of results files (every time step if possible) and include the mesh. If the motion is not coupled to the flow then turn the flow solvers off and just run mesh motion. Then you can see the mesh as it deforms and you should be able to identify what region is causing the problem. Once you know where the problem is you can think of options to fix it.

November 27, 2012, 11:11		#6
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	I just want to clarify that the moving part is separate from the stationary part and is connected using GGI. Because if they were in the same part the mesh would definitely get all tangled up. I think Glenn's idea is best: save results every timestep and see what happens to the mesh. Feel free to post the result here. Just to one iteration per timestep since all you would be trying to see is what is going wrong with the mesh.

November 27, 2012, 12:04		#7
sakalido Member Sara S. Join Date: May 2010 Location: NJ, USA Posts: 41 Rep Power: 16	Chris, The right and left pistons are a separate part and the main domain inside the valve is a separate part as well. The washer in between the right and left pistons is a subdomain inside the main domain. I will try to run the simulations and stop it before it overflows to check the mesh deformation shape. Sara

November 27, 2012, 12:22		#8
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	Instead of trying to stop it you can just have it write an output file every iteration since you said it doesn't do very many. In Pre, go to Output Control > Backup and set output frequency to every iteration.

November 27, 2012, 13:58		#9
sakalido Member Sara S. Join Date: May 2010 Location: NJ, USA Posts: 41 Rep Power: 16	I have four iterations now and I looked at my mesh and tried to make a video of the mesh deformation. I don't see any strange deformation since the movement is very small. Unfortunately the saved video is not helpful to post here bc it doesn't show the mesh, it just shows a blank solid surface! I also looked at the mesh deformation contours, the maximum deformation is 2^-6. How can I check whether I have tangled mesh?

November 27, 2012, 14:19		#10
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	I think you should be able to visually see if there are problems. In CFD Post experiment with showing the surface mesh of various parts and look for anything out of the ordinary. There are also various variables you can plot related to the mesh (aspect ratio, face angle, etc.). Sounds like your maximum deformation is not that big so far, so maybe there is some inconsistency in the way the mesh motion is specified?? Maybe you should post more details about how you have set that up.

November 27, 2012, 16:16		#12
cdegroot Senior Member Chris DeGroot Join Date: Nov 2011 Location: Canada Posts: 414 Rep Power: 18	That image helps explains things for sure. I have questions still about the way the motion is specified on the regions with specified motion. Is the displacement set to be relative to the initial mesh? I think that is what you intend. Are you using the specified displacement option on those faces? Double check that both sides of the interface are specified in a consistent way so that they move in the same direction, not opposite each other. I also wonder if the "affected wall" should be stationary. Seems to me like a rigid part.