I’m working on trying to model a simple I-beam with added plates along the web to analyse their stresses (image below). I’m wondering how it would work when assembling the model in Karamba3d? I’ve seen guides for combining beams and shells (ex: including beam vertices as points in the shell mesh) but that’s usually for large shell elements like floor slabs supported by several beams. Is there a way to combine a LineToBeam element with a small mesh like the one pictured? Or should I recreate the beam as a composite of three shells (flanges + web modelled as shells) and include the plate vertices in those shells? Any tips greatly appreciated!
PS. I noticed the modelled beam component doesn’t visualize the flange radius for the beam despite it being part of the cross-section input. Am I missing something or is that value only used to compute the mechanical properties of the beam internally without affecting the visual model? (note: my rhino file is in meters, to explain the calculations for x,y,z values in constructing the points for mesh vertices)
Web stiffeners are typically used at locations where significant shear forces act on a beam—such as near supports or where a large concentrated load is applied. Their purpose is to prevent the web from buckling.
In your case I would model the beam and the stiffeners with shell elements and activate second order theory analysis. Use the Karamba3D mesher to ensure that the meshes are connected at their vertices. If you enable the WIP components you could also test the new mesher which supports local refinement around the stiffeners.
Fillets of rolled I-sections are not included in the cross-section mesh, but they are accounted for in the mechanical cross‑section properties.
Thanks for the reply!
I’m going ahead with modeling as shells. By the karamba mesher you mean the MeshBreps component? The link takes to the same 2nd order analysis page linked previously.
Unfortunately I’m on Mac so I’m using the older version 2.2; should v3 be released for Mac I’d be glad to try out WIP components.
Since I’m recreating the I-beam using shells, the fillet is no longer considered in the mechanical properties unless I model it precisely, though their contribution is negligible for my purposes. I’m working on optimising the stiffener element for a WAAM production process. At the moment I’m just trying to recreate a simple experimental condition of a concentrated load in the middle of a beam, once I get the process down I can adapt it to different configurations that are more realistic.
If I have future questions about this project I’ll continue in this thread! Thanks for the assistance, I’m a big fan of the product so far.
I am sorry that we still haven’t updated the native Mac version of Karamba3D. The problem is linked to the fact that the intel math libraries do not work there any more. In the Mac version the new mesher is not available.
I’ve run into a roadblock trying to model force flow lines on the rigidity plates. I’ve input their source points and direction (downward Z vector) but I keep getting odd results. Is there a way to tell the Line Results component to only show results for specific elements of the model based on element IDs? How else could I visualise the force flow for a single element of the model instead of the whole thing? I thought about creating a separate model just for the rigidity plates and trying to recreate the effects of the I-beam system with loads and supports but I feel there must be a better solution.
the issue is that in your definition the web stiffeners are currently connected to the beam only at their four corner nodes. In Karamba3D, elements are connected only when their nodes coincide. If you want to rely on the old Karamba3D mesher, you’ll need to add inclusion points wherever the web and the stiffeners should be linked. With the new WIP mesher, you can instead refine the mesh directly along the connection lines.
To model the web accurately, make sure to have at least six elements along its height.
I’ve updated the model as described; I separated the elements (flanges, web, plates) into separate Mesh Breps components, each with unique mesh resolutions. I also stiched the shells together with all vertices present on the seams. Initially this didn’t solve the visualization of line results, but I was able to obtain these results (image) by including all mesh vertices as sources for the LineResults component (I added a CullPattern to take every other point and lighten the computation as the results are near identical when using all of the points). Just wanted to check if this seems like a correct solution since in the tutorial / example files the sources for LineResults are generally just the edge vertices.
GH attached below; sorry for the mess. Appreciate any tips to clean up the file as well.
I would take care of modelling your support correctly? Currently you have large stress concentrations here, what does the desired support actually look like?
The web triangles are a bit narrow, could lead to inaccuarate results. I am also using macOS version of Karamba and usually stick to the TriRemesh mesher for my FE models for more equilateral triangles
for forum scripts, I would always reduce the script to the minimum possible. Also, your components are very dense and it is hard to read the data flow
I don’t know what the purpose of the standard LineResults is. I usually use the ShellVecResults, as the density of stress lines is arbitrary and has no meaning and one cant understand the stress values. It only gives you (a potentially misleading) impression of the force flow…
Also, I am curious, what are you trying to achieve with the WAAM for the I-beam?
Best wishes,
Rudi
p.s.: here a video of the way I am visualizing the principal stresses
Thank you very much for your input! To answer some of your questions:
I modelled the connection this way to reflect a real-life condition of a bolted beam. I’m considering moving the supports closer together to the web’s midpoint, or even moving them inwards along the web. Or, I could go the other direction and simplify with 1 support on each side to create a simple, determinate structure. I’ll refine the loading and support conditions as I move forward with this project now that I’m understanding how to set up the model and obtain results.
Do you suggest adding the TriRemesh before or after Karamba’s MeshBreps? In the latter case I assume I’d need to add the necessary inclusion points (seams between meshes, load+support nodes) into the Features input?
I’ll definitely work on decluttering the script and spacing it out to read the flow better, especially when sharing files on these forums:)
I just did a quick test to compare; my first guess is that LineResults connects similarly-oriented vector results into continuous lines, but not sure how it treats different values for these vectors. Maybe that’s where the lines get denser, to reflect higher stresses.
The research is based on the linked article about using WAAM to produce stiffeners for I-beams. I’m working on optimizing the stiffener’s topology to further reduce material and take better advantage of WAAM’s capabilities.
The video visualization is really cool! What are the points that move along the bottom? Looks like supports based on the visible graphics. Also, are the vectors colored based on their intensity or direction? Would definitely love to develop a similar visualization for this project, perhaps with moving positions for the stiffener plates.
the struggle with TriRemesh is, that the points have to be in the mesh already so that they can be kept as a feature, so I am currently using MeshBreps to include the points and then I am using TriRemesh, a bit cumbersome but works
Cool, the density of the line results depends on the source pts and how the stress field looks like, as far as I understand, the resulting density is therefore not quantitatively useful
In my visualization the colors depict the principal stress values
and yes, I am moving the supports to visualize the impact
Very interesting WAAM application
Furthermore, I think the stiffeners also increase buckling performance of the beam, but this is just my intuition
