Slabless Stairs Design - Modelling & Meshing

[Anonymous]

Hi, 

I am currently working on the design of slabless/sawtooth stair design. I have modelled this in 2 ways, both defined as a shell structure. In the first model, I designed the staircase flights as 1 panel with landing as a separate panel. In the second model, I designed all threads and risers as different panels. I expected to have similar results, but as you can see in attached files, the meshing in the first model is completely disrupted and does not follow the geometry as designed. 

I am experimenting with different ways that allow best meshing possible so I can determine the minimum thickness for the staircase as this would be the main design feature. An explanation as to why this is happening and how best to model it can be of great help!

Thanks in advance,

Luke.

[Simau]

Hi @Anonymous 

For the model "only 3 panels" use Delauney in adnaced meshing option.

[Anonymous]

Hi @Simau 

Thanks for your reply! I was using the "Delaunay and Kang" parameters in my previous tries. Tried using "Delaunay" only as you showed now, however still got an unpleasant result in my model as per screenshot.

Thanks,

Luke.

[Simau]

No problem here

look at what i get

 Try using 0.05 m size

[Rafacascudo]

hi @Simau ,

That is really quite interesting!

I never thought that is was possible to define a panel like that and get a perfect mesh . In my experience ,a zigzag panel with consecutive  ortogonal planes , the result mesh was always what was displayed on @Anonymous  original model. A weird mesh with Robot trying to build a curved mesh to best represent the input geometry.

Even if you try to get back to the original mesh size(15cm) or change to Conns method , the mesh will remain perfect!!!. And running the model will give no errors at all!!!!

Maybe it is because it was created as a "Floor" object , not as a regular general panel.

Anyway , results on the 2 models with the same mesh size and method are quite different!!.

I would stick with the multiple panels model . I really do not trust 3 panels model although it seems to be perfectly correct with no warnings or errors messages.

If it is correct ,Could we then create an entire core wall with a rectangular section using just a single panel???

I don´t think so.

I would like to hear from official support @Rafal.Gaweda @Artur.Kosakowski  @Krzysztof_Wasik what is the official RSA opinion about this

[Anonymous]

HI all,

Thanks for your response @Rafacascudo 

I continued to work on the model and managed to get a correct meshing as suggested by @Simau. I do believe that there were some advances in meshing as versions of the programs were developed. I always found it convenient to stick to my original version, but decided to update to the most recent software and meshing was done differently and function in both models.

However, there still seems to be a discrepancy in the analysis as you say! See the attached screenshots, which shows a panel cut with the BM diagrams. Clearly, results are different. In the 3-panel model, no moment is shown in the thread regions, which seems strange - risers and threads in such a construction are rigidly fixed, and there should be moment transfer. In the various panel model, the moment is in fact shown in both the threads and risers. To my knowledge, the link between two separate panels acts as a rigid connection, as long as their edges are drawn accurately on the same line. Therefore, in theory, both models should give similar results, no? 

The maximum moment is different, and I'm trying to optimize so I can get my structure in line with the design for a 125mm thick slabless stair panel. I need to ensure that the model is working fine in order to do this.

Thanks once again for the support!

Luke. 

[Simau]

@Anonymous @Rafacascudo 

In my messages, I think it was by chance.

With this similar model it is impossible to mesh the single panel.

See model

[Anonymous]

@Simau Very interesting...

Does this mean that the moments given in the 3-panel model are not realistic? And therefore, the whole staircase should be modelled as separate panels for all threads and risers?

Thanks,

Luke.

[Simau]

@Anonymous 

Does this mean that the moments given in the 3-panel model are not realistic?

I don't think so, for me the results with 3 panels are more realistic. Calculation by hand gives a bending moment about 17 kNm to compare with the moment shape and value 17.38 kNm.

Moreover, for the single panel the mesh is weird

[Rafacascudo]

@Simau  I think  @Anonymous  is talking about HIS 3 panels model that, as I said before ,I would not trust

[Simau]

@Rafacascudo 

Just to be clear, what is your opinion about the example on message 9?

[Krzysztof_Wasik]

Hi @Simau @Anonymous @Rafacascudo 

I have remeshed model with 3 panels in versions 2021 and 2019 and model looks as expected.

before

after

So in this case reasonable mesh can be obtained.

However I would recommend using single panels for such stairs definition. In my opinion it will be easier to control loads and results when model will be divided into separate panels.

Problems with results display as described in post 6 is caused by direction selection, while cut definition. For results presentation I would recommend to select local cut direction instead of automatic one related to panel local system of coordinates.  

In model with separate panels, z local axes have different orientation. In that case smoothing option will not work well. Change z orientation using  Geometry >Properties>Local Panel direction or switch off smoothing while results presentation.

 Regarding model test- one panel version. Mesh will be reasonable when small vertical panel part will be added (otherwise Coons meshing method builds incorrect surface geometry)

Results for cuts are the same when presented in local cut direction.

[Rafacascudo]

@Simau  , I managed to get the exact same mesh ,in your test model, for both single panel and 3 panels using Coons method.

And as a consequence , all types of results are also identical

I tried an U shape wall panel with every  mesh settings possible , but this time , the program generated the expected curved surface mesh

So , right now , I am kind of lost.

[Krzysztof_Wasik]

HI @Rafacascudo 

Dummy panel definition will help in this case. 

But generally contours defined in that way are unpredictable. . When I added point in the middle of central or side walls the following geometry is recognized.

So I would divide model in three flat panels

[Anonymous]

@Simau @Rafacascudo @Krzysztof_Wasik 

Thanks for your help! I worked on the inclined plane model by @Simau and managed to get similar results as @Rafacascudo with Coons meshing. A disrupted mesh was what I got initially, however, I solved this when modifying the forcing ratio from "any" to "forced". This way the mesh followed closely the panel as modelled - I believe the extra panel on the bottom edge as suggested in reply 12, would not be required (correct me if I'm missing something! @Krzysztof_Wasik ). The below are the shots I am referring too, and my subsequent results:

With regards to the slabless stairs model, I worked on @Krzysztof_Wasik  suggestions. I hope I understood the local axis system correctly - I selected all the risers (i.e. vertical panels) and changed the local Z axis sense of all of them. In this case, I kept global smoothing on in the panel cut options. Then I tried keeping the original z axis sense for all risers, but changed to "no smoothing". In both cases the system of local axis is "Local (of the cut)" and not "automatic". The results still differed:

Diff Panel - Z not reversed - no smooth

Diff Panel - Z reversed - global smoothing

Thanks,

Luke.