2D-3D Warpage differences

[Anonymous]

Hello everyone, I'm doing a Fill+Pack+Warp analysis on a part with thickness of 2.5mm and ribs of 0.8mm ; the material used is a PP (Borealis MD231U).

I first chose a Dual Domain analysis, as the mesh match is 93% and reciprocal is 94%, moreover this part could be considered a thin walled one.

The resulting deflection in Z axis is -7.5 / 15.3, which is quite relevant.

Then I decided to try a 3D analysis to compare the results, using the same process parameters; in this case the resulting deflection in Z axis is -1.5 / 1, really very different to the Dual domain ones.

Which one do I have to trust? Any suggestion for the best option to check (small/large deflection, mesh aggregation, etc..) ?

Thanks in advance

Alessio

[bernor_mf]

Hi Alessio,

here is some information in an article:

Why is warpage direction different between DD and 3D mesh types

Hope this helps.

Regards,

Berndt

[Anonymous]

Thanks for the answer, Berndt.

I had already checked the link you provided and reviewing the causes as written in that article, I'll give my answers:

"Causes:

Typical reasons: 
1) The packing profile is set to use % Filling pressure vs time, using 10s/80% of injection pressure, 
The injection pressure is then different between the DD and 3D mesh type part causing a packing pressure difference, causing the warp deflection difference. 

2) The part is a typical thin wall part, and there is not a 1:1 mesh relation of converting the Dual Domain mesh to a 3D mesh type. 
Hence creating a too large injection pressure difference between the DD and 3D mesh type part. "

1) In my analysis, I used "Packing pressure vs. time" and the values are the same.

2) The pressure at switchover is 17 MPa for 2D and 19,5 MPa for 3D, so I think they are comparable, moreover the number of layer in MF 2018 is set to 10, and I refined the tetra mesh near gates.

Could be there other causes?

Thanks

Alessio

[bernor_mf]

Hi Alessio,

good you already looked this up so modeling is as expected.

Is packing time sufficient for both studies?

And, if you compare volumetric shrinkage? (at ejection, and averaged for 3D?

Is weight result similar?

Material Data used is:

MD231U : Borealis Europe

The material has generic Mechanical Properties.

(As the headlines in material properties tab Mechanical Properties are red.)

There is no shrinkage data, which is used for CRIMS in Midplane/Dual Domain mesh types and Warp to enhance result prediction.

ED230HP : Borealis Europe has measured data, and is a similar grade.

Maybe give a try on both mesh types to see if this is material dependent?

As PP, did you also ran Cool?

Regards,

Berndt

[Anonymous]

Hi Berndt,

here attached the volumetric shrinkage plot.

I'm not so confident with the result "Average Vol. Shrinkage" and I prefer to check the Volumetric Shrinkage at the end of the cycle.

I think the packing phase duration is enough because the frozen layer fraction is 1 for the 95% of the part (layer not frozen for a small part near gates).

Ribs are overpacked, as in the other analyses.

Part weight differs of about 0.6g between 2D and 3D.

I tested the ED230HP as suggested, in 2D the deflection referred to Z axis is -7 / 14, same behaviour of the MD231U.

I think I'll go for the worst case, choosing the 2D for the report.

I

[Anonymous]

Another thing I noticed: whereas the parameters are the same, with the 2D I have an extimated linear shrinkage of about 0,85 % while with a 3D analysis the shrinkage is about 0.65 % . Beeing the material a 20% Talc filled, I assume that the 2D could be more affordable...

I really don't now what to consider for the report 😞

[bernor_mf]

Hi Alessio,
first, 3D average volumetric shrinkage plot is just to get an average to compare with a volumetric shrinkage for Dual Domain mesh type.
Fully aware of the limitations of the 3D average volumetric shrinkage plot.

Further, in matter of trend I believe Dual Domain mesh type looks most as expected.
As the thinner ribs are packed during filling and freeze off, they will not shrink as much as general thickness of part.
Hence it will bow as in picture provided in previous post.

The magnitude could indicate an unstable warpage, hence maybe a buckling analysis should be considered.
Think you should run a buckling analysis in Midplane mesh type, maybe 3D (with two values, but pretty slow in 3D).

The estimated shrinkage in Deflection properties is based on overall shrinkage.
As DD mesh type warp has higher magnitude, it will show higher shrinkage.
This article explains the background.
How Autodesk Moldflow Synergy determines the Automatic shrinkage compensation percentages

Does cavity weight result plot show the same?

Regards,

Berndt

[Anonymous]

Hi Berndt,

regarding the part weight, it differs of about 0.6g between 2D and 3D.

I'm sorry, I checked the 3D log just now and I found some warnings: 

------------------------------------------------------------------------------------------------------------------------------------------

** WARNING 304920 ** There is insufficient refinement of the tetrahedral mesh in some areas, which
may affect solution accuracy. Inspect with the "Node layer number" plot and
consider improving the mesh. [For the 10 tetrahedral refinement layers requested,
the node layer number is expected to reach 6 at the part centerline.]
** WARNING 304930 ** The node layer number on the part centerline is 5 or less for 52.3 percent of the part.
** WARNING 304940 ** The node layer number on the part centerline is more than one less than requested
(i.e. 4 or less) for 24.3 percent of the part.

** WARNING 304950 ** The quality of the tetrahedral mesh used for the polymer is poor, as described
in the warning/s below. This may affect solution accuracy. Consider improving
the tetrahedral portion of the mesh using the Mesh Repair tools, or even remeshing.
** WARNING 304980 ** The aspect ratio of the tetrahedral elements used for the polymer exceeds
the recommended maximum of 30 in 17.3 percent of the mesh.

---------------------------------------------------------------------------------------------------------------------------------------------

Probably, even if in the settings it was given 10 as layer number in thickness, the starting mesh size is too big to sufficiently refine the tetras.

I'm going to remesh the part paying attention to the mesh quality and then I'll let you know.

For now I'll go with the 2D.

[bernor_mf]

Hi Alessio,

right, that could be the case.

For a 3D mesh type, the general rule of thumb is a tetra edge length on surface 1x to 2x the thickness.

The more layers of tetras, the finer surface mesh needs to be, otherwise squashed tetras with high aspect ratio.

Also when thin ribs, the tetras will be more squeezed in layers to fit in the thickness.

For your case maybe edge length 1x thickness and 8 layers would be sufficient for a test?

Regards,

Berndt

Moldflow: Recommended mesh size for different mesh types

[Anonymous]

 I used a surface edge lenght of 4 with a part thickness of 2.5 , so a ratio of 1.6 ; clearly it's not enough.

I'll go for 2mm and after the refinement, I'll let you know, then I could accept this as a solution 🙂

Thanks for the time spent helping me!!

Kind regards

Alessio

[Javier.Jubierre]

Hi @Anonymous,

Converting DD meshes in 3D meshes - "to check results" - by running the second step of the 3D mesher, is a very common mistake. DD and 3D warp models are based in different mathematical approaches, and "re-using" the DD mesh is a mistake to avoid.

In this case you have a material without CRIMS, and a part that looks quite standard for DD. So, I would search (as already suggested) a similar material with CRIMS or remesh completelly the part.

In 3D you have a plot called "Node Layer Number". This plot should show you a value greater than 6 to be sure you already allocated 8 layers. If you get less layers the warp physics will not work properly on your simulated part.

I hope it helps.

Cheers,

Javier

______________________________________________________________

If my post answers your question, please click the "Accept as Solution" button. This helps everyone find answers more quickly!

[bernor_mf]

Hi Alessio,

no worries, happy to help. 

Cannot see on previous provided pictures how gate for 3D is modeled/meshed.

Just wanted to add:

For gate beam attached to 3D mesh type, it is recommended that the beam shadows at least four tetra faces.

I also use the minimum 3 beam elements to represent the gate, many times 4 or 5 elements.

If you still experience a too big difference, Autodesk tech support and/or R&D probably need to review this.

Regards,

Berndt

[harald_goetz]

Hello @Javier.Jubierre,

you said:

Converting DD meshes in 3D meshes - "to check results" - by running the second step of the 3D mesher, is a very common mistake.

Really?

There is an Autodesk University Video

http://au.autodesk.com/au-online/classes-on-demand/class-catalog/classes/year-2014/moldflow-insight/...

as well as this here

https://www.youtube.com/watch?v=hkqEefht4xY&t=909s

from the "Built your simulation IQ...

Both are rather old, but 100% Controverse to what you mentioned.

Can you please refer in detail why this is an mistake?

I'll just try to avoid this fatal (?) error in the future, but better to know why?, right?

Cheers

Harald

[bernor_mf]

Hi,

@Javier.Jubierre

I agree with Harald.

From what I have learned surface mesh is prerequisite for 3D tetra mesh.

As referered by @harald_goetz from AU2014 and referred

SM5706-P: Purging Old Misconceptions: Best Practices for Meshing in Simulation Moldflow Insight
in presentation pdf p.27 it is stated:

"Dual Domain is required first before meshing to 3D
- Not required but recommendened and highly beneficial
- Primary advantage is easier access to address troubleshooting
- With native CAD less beneficial as fewer model related issues exist."

So still going from Dual Domain to 3D should be a valid workflow

Here Shoudong, the Moldflow mesher developer, mention a workflow going from 3D to DD, refine surface mesh, and then mesh to 3D.

Meshing Tips: How to Remesh 3D Meshes?

@xusho

Shoudong, can you please shed som light over this, if Autodesk has a new statement for using  DD to 3D mesh?

Thanks.

Regards,

Berndt

[Javier.Jubierre]

 Hi all,

Thanks for your comment, lets see if I can bring a little bit of light on this issue.

I am not saying that meshing first in DD, checking/correcting surface mesh, and meshing in 3D is an error. That was our recommended procesdure until 2017.3 where we launched our automatic mesh sizing and 2018 where we launched the possibility of launching simulations without previous mesh. Actually, we do not recommend manual modification of the Moldflow mesh anymore (This is an official statement of Moldflow, not mine).

What is an error is using a DD mesh of an existing project to check warp results into 3D. The main issue is that DD meshes are coarser than 3D. That means that even defining 6, 8, or 10 layers in the mesher, the final mesh will never allocate so many layers, the elements will be simply too big. Remember, this setting is a goal, not a constrain. Thus, to achive the allocation of so many layers the surface/DD mesh should be much finer/smaller.

In addition, to get valuable results of warp in 3D we need to have 10 or 12 layers of elements at least. (Therefore we changed the default number of default layers from 6 to 10). Due to classical DD parts are thin and long, to generate 3D meshes for such parts involve a huge number of elements, which make the simulation quite time-consuming, and therefore  do not recommend them.

I guess we agree that this dicussion is not needed for filling and pack results, isn't it?

I hope it helps,

Javier

______________________________________________________________

If my post answers your question, please click the "Accept as Solution" button. This helps everyone find answers more quickly!

[Anonymous]

Thanks for your comment, Javier.

The discussion is about 2D/3D warpage comparison, I was looking for a reason of such difference among theese results.

I still don't understand your suggestion, you're saying that for this kind of part with 2.5mm thickness the DD mesh is more affordable because the 3D would never reach a sufficient number of layer in meshing without exceeding element's number?

Thanks in advance

Alessio