Stress Linearization query

Feature Requests
1

Hi everyone,

I have seen some previous messages about this topic (applied to vessel design). I want to know if, according to your experience, could recomend me if the next procedure could be developed on Prepomax, and is consistent with the requirements of ASME?

1- Modelling with shell elements

2- Calculation with Calculix

3- Evaluation of stresses on nodes. Because of the expansion of shell elements into solid elements, could be obtained 2 or 3 points through the thickness.

4- Code checks (I have found this excel, seems to be very useful)

2

Just keeping in mind that CalculiX has “fake” shell elements. So it’s more like using solids according to codes.

You can always define a composite shell section, even if all layers are of the same material. Then you will have more points with the results.

Actual stress linearization procedure is not yet implemented in PrePoMax, so it’s still a valid feature request. As workarounds, you can try other software such as ParaView or FreeCAD (needs some fixes, though).

Currently, you can only switch between membrane and membrane + bending by changing the OUTPUT=2D/3D parameter for field output requests.

2 Likes
3

I think 2-3 nodes through the material thickness are not enough to capture the real stress field. Usually, a solid mesh with multiple elements through thickness is required - maybe a multi-layered shell.

2 Likes
4

Just to add something from the code side ASME VIII Div. 2, in the Annex 5-A.4, the stress Integration Method actually treats solids and shells differently:

  • In continuum (solid 2D/3D) elements the stresses are linearized along a stress classification line (SCL) through the thickness, so you need enough through-thickness results for the integration to be meaningful.

    image
    image1205×812 152 KB
    image
    image1191×425 52.4 KB

  • For shell elements (annex 5-A.4.2): membrane and bending stresses are obtained directly from the shell stress resultants (forces and moments), so no through-thickness linearization is needed. Peak stresses are then handled separately.

    image
    image1216×733 79.3 KB

So the number of points through the thickness really matters for the solid approach, not so much for the shell one. With CalculiX, though, since shells are expanded into 3D elements, strictly speaking you’d be following the solid procedure, unless you can extract the top and bottom stress tensors and work with those directly.

For solids it’s a different story, since you need to define an SCL and integrate the stress tensor along it. There I usually use the built-in tools in ansys or abaqus, and when I don’t have them at hand, I export the results (.dat) and work through it with some Python routines, similar to the Excel sheet shared in the post. Having stress linearization available in PrePoMax would be really useful for this kind of work.

2 Likes
5

Yes, that’s what I’ve mentioned above. Currently, CalculiX has only one true shell element and it can provide SPOS/SNEG stress, but it’s a linear triangle (S3) and limited to linear (small strain) analyses.

Stress linearization for solids is currently only possible with 3rd party tools. FreeCAD has it, but it’s flawed. ParaView has only plotting along line, but it could be extended to perform stress linearization with some custom scripts.

1 Like
6

Hi @FEAnalyst thank you for your answer. I think that orient future development of software to practical cases would be very interesting. For example, I am usually involved on structural analysis of bolted and welded connections submited, so analysis such as fatigue, stress linearization, etc. would be very useful for me. And I think that a great part of the comunity is involved on similar analysis.

1 Like
7

Actually, connections are a weak side of CalculiX, but PrePoMax tries to improve that and already has new tools to handle joints.

Of course, it’s always easier to add purely pre-/postprocessing tools than solver algorithms so things like stress linearization should be doable.

Stress linearization is often used in DBA, particularly in American codes, but other ones make use of it too.

1 Like
8

Multi-layer definition, is it implemented yet on Prepomax, or shall be defined on the Calculix?

9

It needs to be defined using Keyword Editor:

*SHELL SECTION, COMPOSITE, ELSET=elset_name
t_1, ,material_name
t_2, ,material_name
...

And you need S6 or S8R elements for that.

2 Likes
11

Thank you very much!

12

Check out this post at the CCX forum:

3 Likes
13

Thank you very much Sergio. I hope this could be implemented into Prepomax as soon as possible.

It is impressive the velocity that today applications are developed.

As you have the results yet on both apps, could you check what results you obtain if you include the table into an AI app (ChatGPT, Gemini, Claude, Copilot..?).

Thank you.

14

That new postprocessor is already (at least mostly) generated by Claude.

15

If your aim is “validate” the results, in the post I have attached a link to the report made by Pveng using Abacus. If you look also they have made a report with SolidWorks simulation and the values doesn’t agree 100%. Today later I will look in my files because I have made an Excel comparing those two reports and Mecway, that gives exactly the same results as CCXview.

16

About shell results, the solver FEMaster works with real shell elements (that was my initial point of interest in the solver), and they results can be readed with CCXview (didn’t try yet with shells specifically). I have never worked with shells with parts under ASME requirements, could you share some reference report example?

Thanks in advance.

1 Like
17

Hi @SergioP1975 I mean that if you give a table to some application and you ask about linearization, what results can give you. It could be a way to do on the future and with applications as Prepomax, that today hasn’t implemented it yet.

18

It is also what I am looking for, a study with shell elements and not solid. If you find, please share :slight_smile:

19

Look in the Pveng site, they have lot of well documented cases, but not remember now if they work with shells.

1 Like
20

I don’t see any problem about that, if it is well implemented and checked, I have no problem about that. I prefer that than commercial softwares that don’t have the intention to develop the requested capabilities by users.

21

My point was mainly that we pretty much already have stress linearization results from Claude this way. I don’t mind using AI as a support for coding either. Of course, everything has to be done reasonably - well checked and not just 100% AI generated without a careful review.