CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Help needed on compression modelling!! (http://www.cfd-online.com/Forums/cfx/79376-help-needed-compression-modelling.html)

 Kaustatic August 22, 2010 11:33

Help needed on compression modelling!!

Hi, I am trying to model a complicated multiphase process using CFX programmimg, to which I am very current. I have started working on it by making it a combination of two concentric cylinders in which one on top is a 30 mm wide , 20 mm long cylinder and the lower one starting from the end of the broader is 6 mm wide.

Keeping the narrow cylindrical region constant (stationary as outlet region), mesh displacement has been attempted on the broad cylinder on top. the top flat layer (30 mm wide) and the wall of the top layer has been subjected to a constant velocity. (The fluid is water at 25oC). I tried to set up a transient by trying to displace the top mesh by 5 mm following the CFX tutorial on fluid interaction and mesh deformation. A direct application showed only the initial flow state and the final flow state(right at the first timestep) and the rest of the time steps the mesh died out.

I tried to right expressions for the displacement, by multiplying a given velocity with the timestep and applied it to the top surface and the side qwall of top cylinder. but now, no movement at all. Help!!

 ghorrocks August 22, 2010 18:34

I have no idea what you are trying to do. Can you post a drawing of it?

 Kaustatic August 24, 2010 05:08

2 Attachment(s)
Hi thanks for your reply. I am putting up the two states of the system. The first step has the upper cylinder (water, 20.7 mm high) at full length. Next under the boundary conditions, the cylinder has been compressed by 5 mm. The problem I am facing is, although I am setting up a transient, the complete movement is occurring at a single time step and each step is showing <Final>. For the subsequent time steps, the mesh can be seen to die out.

Please help me out with the transient, I want to see it moving in steps as the height reduces under apoplied velocity profile (ideally a force)

 ghorrocks August 24, 2010 06:34

You need to define a function which describes the motion of the moving bit as a function of time. displacement=velocity*t or something like that, and you define the velocity.

 Kaustatic August 25, 2010 18:34

i have tried that as well: taking the initial height as HOld, the inital velocity as velold, the mass calculated from density and volume of the top cylonder as m, assuming an applied force (force_z(mesh....)), F, and timeStep, the new height after the total time is,

HNew = HOld - (velOld * tStep) - Fm/tStep^2.

but it ended in error due to negative volume. :(

if I ignore the force, it shows no movement...(or is it the movement is small? I was taking ttotal = 7.5e-3 sec, and tStep = 5e-5 sec

 ghorrocks August 26, 2010 06:41

Debugging mesh movement is always tricky. A good way to start is to turn off all the solvers using expert parameters, make it output a result file every time step and run the simulation. It will then run fast (as it is only solving the mesh movement) and show you what the mesh looks like every time step and that is the first step to debugging it.

Also have a look in the tutorial examples for mesh movement. They show you how to set it up.

 Kaustatic August 27, 2010 08:00

Regarding your suggestion to study mesh movement!!!

I am sorry but could you please let me know the steps for the mesh-only movement test that you suggested? as I am new to this field (only couple of months with not a long time in hand for this project) a little more experienced help would be greatly appreciated.

 vmlxb6 August 27, 2010 15:41

Mesh motion

Go to CFX Pre-Insert-Solver-Expert parameter and disable everything.

 ghorrocks August 28, 2010 06:41

As Eric said, in CFX-Pre go to insert-solver-expert parameters and I think you are looking for the solver overrides. Turn everything off except the mesh motion solver. Then go to the output tab and make it save a full results file each timestep (which includes the mesh). Let it run, then load it in CFD-Post and see if the motion generated is what you intend.

 vmlxb6 August 28, 2010 15:56

@ Ghorrocks

@ Ghorrocks:

I did not find the mesh motion solver in the expert parameter-> model over-rides. Are you talking about the expert parameter-> discretisation tab-> meshdisp diffusion scheme ?????

Thank you for all your help.

Eric

 Kaustatic August 29, 2010 01:38

'compression modelling'

I checked the path you suggested. But everything was already unchecked there, so I was looking into mesh motion.

 Kaustatic August 29, 2010 06:02

compression modelling

I checked at every timestep...there is no motion beyond the first timestep.
Is it somthing to do with transient initialisation at each iteration? I wrote the expressions as:

velocity = 1.67 mm/sec
tTotal = 3 sec
tStep = 0.1 sec
displacement = velocity * tStep

this displacement I applied to the top surface and made the wall mesh movement Unspecified. The result seems like the movement is initialised to 0 each time and one stepo thereafter so the same result is seen. how can I add a changing initial height of cylinder at each iteration?

 vmlxb6 August 29, 2010 06:29

@ Kaustatic

The problem seems to be in the UDF written, that is the CEL expression. I am attaching the UDF written for my case which is the flow over a cylinder. Do not mention the wall which is moving as unspecified but set the mesh type as Specified displacement. If you want the wall to move only in the x direction (say) then
Mesh motion> X component ----------> displacement (as mentioned in the UDF

Mesh motion> Y component------------> 0 [m]

Mesh motion> Z component------------> 0 [m]

Check out Chapter 22: Fluid structure interaction and mesh deformation tutorial.

Hope this helps.

Regards,

Eric.

P.S: Did you solve the negative mesh volume error ??? Can you tell me how you checked the mesh deformation at every time step ???

 Kaustatic August 29, 2010 09:37

compression modelling

Hi, thanks for the detailed guideline actually I am already working in that direction and the whole cylinder is moving. the negative-volume error is really down to the physics. You need to check whether:

(a) the physics is pulling the structure opposite to the intended direction i.e. say if the volume is to the +z (in my case), its pulling it in -z. Or some mathematics in the physics is producing negative value;

(b) the wrong volume element is made to move.

But my problem is I have been getting the final movement in a single timestep. Now I have tailored it as
displacement = areaAve(Total Mesh Displacement Z)@upperwall + (velocity *tStep)

but now the movements each time step is very small...I dunno how to make it significant.

 Kaustatic August 29, 2010 19:41

Thanks

I got the present problem solved I think folks!! Thanks for all the suggestions!

 ghorrocks August 29, 2010 23:29

Your expression just evaluates to a constant. You need to make it a function of time. Try displacement = velocity * t. "t" is the variable time.

And please do not private message me again unless the message is indeed private. If it is a question on this thread then post it here on the forum.

 olivlly May 29, 2013 08:20

Hi,can you give me some suggestion on the similar case

4 Attachment(s)
Quote:
 Originally Posted by ghorrocks (Post 273292) Your expression just evaluates to a constant. You need to make it a function of time. Try displacement = velocity * t. "t" is the variable time. And please do not private message me again unless the message is indeed private. If it is a question on this thread then post it here on the forum.
can you help me to find what problems i have

I am trying to simulate a case using CFX programmimg.
A 2D rectangle vessel with infinite length, filled with water, and my purpose is to check flow field when the left wall moves towards the right wall at a certain velocity.

what i have done is build the mesh(6m long, 1m wide and 0.02m thick as CFX can not read 2d mesh).then read the mesh in CFX, see attachment 1 Attachment 22223. the BCs I set is:

front and back faces --->symmetry ( 2d simulation)

top and down faces--- >opening (see attached "opening bc")Attachment 22224 the reason i think is that opening BC can let fluid out and in, sth like infinite length

left face-->wall (mesh motion--> specified displacement; Mesh motion> X component ----> disp; y component ----> 0;z component ---->0 )
right face-->wall (no slip wall)

domain setting see attached "domain setting"
Attachment 22225

configuration setting see attached "configuration setting" Attachment 22226

CEL i edit was:
disp = disp desired-disp mesh reinit
disp desired = 10[mm]*t/tTotal
disp mesh reinit = 10[mm]*Mesh Initialisation Time/tTotal
ort = minVal(Orthogonality Angle)@REGION:WALL<30[deg]
tStep = 1e-2[s]
tTotal = 1.0[s]

the problem is I can run it 1s, it can give result without problem , but the move wall (left face) is still at its original position.
when I changed the tTotal to 2s or greater, CFX said that negative mesh volume or invalid mesh or mesh fold.

is there any problem with my BC setting ?can you give me any suggestion? if i do not simulate the vessel with infinite length but a certain length, and i just want to simulate the flow field in the case the left wall moves towards the right wall, how should i do？ thank you very much!

 ghorrocks May 29, 2013 19:15

This sounds like a problem with your mesh motion. Do the following:
* turn the flow, multiphase and any other fluid solvers off with expert parameters (solve fluids=f and all the others).
* set it to output a transient results file at each iteration including the mesh.
* Run your simulation. It should run really quick as it is only doing the mesh motion.
* Carefully look at the mesh motion in the post processir, especially in the bit just before it crashed. You should be able to see the bit of mesh which is folding.
* Now you know what area to fix!

 ghorrocks May 29, 2013 19:15

This sounds like a problem with your mesh motion. Do the following:
* turn the flow, multiphase and any other fluid solvers off with expert parameters (solve fluids=f and all the others).
* set it to output a transient results file at each iteration including the mesh.
* Run your simulation. It should run really quick as it is only doing the mesh motion.
* Carefully look at the mesh motion in the post processir, especially in the bit just before it crashed. You should be able to see the bit of mesh which is folding.
* Now you know what area to fix!

 olivlly May 29, 2013 23:22

Quote:
 Originally Posted by ghorrocks (Post 430857) This sounds like a problem with your mesh motion. Do the following: * turn the flow, multiphase and any other fluid solvers off with expert parameters (solve fluids=f and all the others). * set it to output a transient results file at each iteration including the mesh. * Run your simulation. It should run really quick as it is only doing the mesh motion. * Carefully look at the mesh motion in the post processir, especially in the bit just before it crashed. You should be able to see the bit of mesh which is folding. * Now you know what area to fix!
thank you very much!
but sorry, as i am new,can you explain more detailedly
* turn the flow, multiphase and any other fluid solvers off with expert parameters (solve fluids=f and all the others).
(do you mean disable everything in expert parameter except
expert parameter-> discretisation tab-> meshdisp diffusion scheme
or expert parameter-> model over-rides ?)

* Carefully look at the mesh motion in the post processir, especially in the bit just before it crashed. You should be able to see the bit of mesh which is folding
(how to check the mesh motion, through animation? )

All times are GMT -4. The time now is 08:33.