Discussion Groups

Simulation Moldflow Insight

Reply
New Member
kathychew
Posts: 1
Registered: ‎12-21-2012

Meshing a component

564 Views, 7 Replies
12-21-2012 04:42 AM

I need to mesh a component with thickness at thin part is 1.5mm while 3.4mm for the thicker part of the component.

 

So,what mesh should i use? Is it i need to mesh with dual domain and then remesh with 3D?

Please use plain text.
Contributor
MartinCavalry
Posts: 21
Registered: ‎03-01-2012

Re: Meshing a component

12-26-2012 07:20 AM in reply to: kathychew

Hi Kathychew,

 

Your wall thicknesses suggest that your part is a thin-walled component.

I think a 3D mesh is more suitable for bulky geometries while the DD mesh is more applicable for thin-walled ones.


I would tend to use the DD, but to make sure if it`s the right one, I recommend you to do the following:

- Using the same mesh density, mesh the part with 3D and DD in separate studies. Look for the difference in computational time and results accuracy. Decide which of the results obtained from 3D and DD make more sense. Ideally you should see no significant difference. Choose your mesh type.

- Perform a mesh sensitivity analysis to check what the optimal mesh density is. Increase the element size until the results do not differ from each other significantly.

 

Hope I could help.

Regards.

Please use plain text.
Mentor
madhukeshwart
Posts: 276
Registered: ‎03-18-2009

Re: Meshing a component

12-26-2012 07:24 PM in reply to: kathychew
If part having wide variation in thickness,
if part having thickness > 3.5 mm,
if mesh match and reciprocal match is < 85%

I recommend to go with 3d mesh..

If not go wish Dual domain.



Thats is the option..


Madhukeshwar Talwar
FIAT-CHRYSLER ATC, Chennai
mail: madhukeshwart@gmail.com
09600060862
======================================
Please use . Accept as Solution and Give Kudos as appropriate to further enhance these forums. Thank you! .....
Please use plain text.
Active Member
ashwinkumarnr
Posts: 10
Registered: ‎03-27-2012

Re: Meshing a component

01-02-2013 09:46 PM in reply to: MartinCavalry

Hi Martin,

 

Can you elaborate on mesh sensitivity analysis.

What is sensitivity analysis, how to do it.

 

Regards,

Ashwin

Please use plain text.
Contributor
MartinCavalry
Posts: 21
Registered: ‎03-01-2012

Re: Meshing a component

01-03-2013 08:07 AM in reply to: ashwinkumarnr

Hi,

 

The major difference between DD and 3D meshes is the way how they obtain the solution.

As far as I know, DD mesh uses the Hele-Shaw solver (non-turbulent Newtonian flow) and makes certain simplifications for thermal conductivity/convection in thickness direction. Also, it ignores the effects of inertia and gravity.

 

The 3D mesh (a full volume mesh which truly represents the part) uses Navier-Stokes solvers and makes fewer assumptions. Heat conduction is calculated through the part and gravity/inertia can be taken into account too. This one can be more computational expensive and recommended for thick parts.

 

Because of the nature of the DD mesh and the Hele-Shaw solver, this is more suitable to parts that have very low thicknesses. In my experience, if the uniform thickness is <=2-3mm one may use the DD, above that, the 3D mesh can be more applicable (especially if you have wide variation in thickness).

 

But, in any cases, the part thickness, flow length and volume should be considered.

 

Besides, if you have a uniform wall thickness of 3mm and the mesh match and reciprocal match is <85% in DD mesh, the solution is not necessarily to switch from DD to 3D but to improve the mesh with respect to quality and density. It is advised to check the mesh by Mesh / Mesh Repair / Mesh repair wizard to determine you have good mesh quality (low aspect ratio, no inverted tetras, no collapsed surfaces, etc). Also, it is important to check for thickness when using DD mesh as the thickness is determined by element matching, in 3D mesh the thickness comes from the CAD model itself.

Repairing the mesh is not always necessary but in terms of solution accuracy and computational time, this is an important step in pre-processing. For example, poor mesh quality (extremely large aspect ratios) can cause excessive computational time with low accuracy, or even, convergence problems.

 

For a quick sensitivity experiment:

 

- Import your file and click on Create / Mesh;

- In Mesh tab / click Generate Mesh. Set the Global edge length (you should see a default value for your part set by MoldFlow);

- Create separate study files and mesh your part with different global edge lengths so to obtain different mesh element sizes (that global edge length should be decreased to increase the element size);

- repair the mesh if it is needed and run the studies.

 

At the end, for example, you can use several output parameters to check for mesh independency. This means that above a certain element size you should see negligible difference when looking at your results. When you reach the independency, there is no need to increase the element size beyond that, hence you can use that specific edge length for your further studies.

 

All together the finer mesh you have the higher accuracy you get (provided your mesh quality is appropriate) but longer computational time is required.

This is why it is important to perform a sensitivity analysis, to find the compromise between solution accuracy and computational time.

 

You could use 3D - and DD meshes in separate studies and run the analysies separately so you can compare not only the accuracy but the difference in results that come from the two different meshes.

 

Also please note, using coarse(r) meshes will maybe provide you reliable fill time or inj. pressure results, but those meshes will unlikely to pick other important features, such as weld/meld line formations, air entrapment.

 

 

I attach an image of the results of a sample analysis.

You can notice that the critical change (accuracy!) in fill time, pressure, clamp F occurs between element size ~20000-30000. Above that, a very little change is observed, although there is a "valley" at ~40000 elements,indicating a slight oscillation in accuracy. Beyond 50000, the curve tends to "flatten out"  so one use ~50000 elements with confidence.

 

Sorry for stretching it too long. Good luck with your studies.

Regards,

Martin

 

 

 

Please use plain text.
Mentor
madhukeshwart
Posts: 276
Registered: ‎03-18-2009

Re: Meshing a component

01-08-2013 07:34 AM in reply to: kathychew
yes, I agree with Martin....Many times I recommend the same.. But lack of time constraints, We will never be able to do it.

Regards,
Madhukeshwar Talwar
Madhukeshwar Talwar
FIAT-CHRYSLER ATC, Chennai
mail: madhukeshwart@gmail.com
09600060862
======================================
Please use . Accept as Solution and Give Kudos as appropriate to further enhance these forums. Thank you! .....
Please use plain text.
Member
J_Liu
Posts: 4
Registered: ‎02-20-2013

Re: Meshing a component

02-20-2013 05:49 AM in reply to: MartinCavalry

Hallo Martin,

i saw ur reply when i searching for meshing. It is quite helpful! Here i have a question i was wondering whether u could help me. I simulated the models with different global edge length from 0.3 to 6mm and the wire sweep, viscosity ..etc. results are quite different from each other.  Does GEL really influence the simulation results so much??  How??

i  thought it is just a meshing. Finer meshing, better accuracy. But it seems not exactly like this.

Thank you for reply!

Greetings,

Jinghan

Please use plain text.
Contributor
MartinCavalry
Posts: 21
Registered: ‎03-01-2012

Re: Meshing a component

02-26-2013 07:04 AM in reply to: J_Liu

Hi Jinghan,

sorry for getting back to you lately. It is difficult to conclude anything as I can`t see your part and mesh, but as for the meshing, have you checked the quality of that?

Also, when you`re doing your meshing studies, please ensure that the process conditions remain the same.

Kindly hope I can help you with this further, but would you be more specific - what do you mean by gel (epoxy?), what is the magnitude of change of your output parameters?

 

Regards,

Martin

Please use plain text.