Torsion and Warping Constants for Built-up Sections 2

[PIEngi]

Hey all,

I have a built-up shape consisting of (4) channels welded together that form a sort of box section, see attached. I am trying to determine the Torsional Warping Constant Cw and Saint-Venant Torsion Constant J for this combined section. These constants are required for finding the bending and compression capacities per AISI S100 code. I have the Cw and J values for each individual channel, but cannot find a way to calculate the combined values. I have searched for similar threads on here, but not exactly finding what I am looking for. I can provide dimensions as needed, but really just looking for a general formula on this. Any help is appreciated. Thanks all.

Edit: I have the ShapeBuilder software, however it won't calculate torsional properties for combined sections, and the drawing tool is not fleshed out enough to create this as one shape easily. I have looked into other software packages, but I have limited access to buying new programs. I can look into any program recommendations/input you provide. Thanks.

 

[KootK]

For this, I feel that it will be reasonable and conservative to calculate J based on the four plate elements that make up the box and to ignore the warping contribution.

 

[Aesur]

When I used to use Shapebuilder to get these properties for custom aluminum profiles I would draw the profile in AutoCAD and region them to make a solid element that could then be imported into Shapebuilder to calculate the properties. Draw the outline of the shape, then region the shape, then draw the outline of the interior open shape and region it and then use the subtract command to remove the interior from the exterior creating the opening. I forget what CAD file type was importable, but believe it may have been dxf.

 

[SJBombero]

I'm not sure why Shapebuilder isn't working. I calculated the torsional properties of the attached combined section the other day. If there are some inherent limitations to Shapebuilder I'd like to know. I'm considering buying it for my business.

https://files.engineering.com/getfi...-b0fc-d0cf6c2c1079&file=Crane_Beam_Report.pdf

 

[Lomarandil]

Sometimes shapebuilder is touchy about how shapes intersect (or don't). What can appear as adjacent shapes that you want to consider compositely might have a small overlap or gap that can cause trouble.

Older versions of shapebuilder also needed you to perform "advanced analysis" before returning the torsional coefficients from FEA.

Robbie has a great python module for computing section properties here:

https://sectionproperties.readthedocs.io/en/latest/

And check out Agent666's blog series about using it for built up members here:

https://engineervsheep.com/tag/section-properties/

 

[PIEngi]

Thanks everyone for your quick replies. For whatever reason, the merge function in ShapeBuilder would not combine all the channel shapes into one section, and so wouldn't run the advanced FEA to get the torsional properties. I ended up getting it to work after fiddling with it for a bit, which finally gave me the Cw and J values. The J value is close to one I had calculated a basic four plate element (thanks KootK for the suggestion), and it ended up spitting out 0.16 in^6 for Cw. Pretty close to 0, good enough for me.

Robbie has a great python module for computing section properties here:

https://sectionproperties.readthedocs.io/en/latest...

Thanks, I had stumbled upon the first link but wasn't sure how it would handle built-up sections. The second link has some great info, and seems like this can handle built-up sections pretty well. I think I'll use this to double check the values I'm getting from ShapeBuilder (and see if it can replace ShapeBuilder for me). Thank you for the link, much appreciated!

If there are some inherent limitations to Shapebuilder I'd like to know. I'm considering buying it for my business.

I've used ShapeBuilder for a few projects now. Overall it is pretty powerful, but very cumbersome to use. I actually didn't know you could import from AutoCAD, thanks Aesur for this nugget. I think drawing in AutoCAD and then importing is probably the way to go for any custom shape. From the ShapeBuilder help file, drawing lines have to snap to a grid vertex, and the dimension cannot be changed after the shape is drawn...I don't see how you can draw complex shapes this way. For standard shapes it works well enough. I think RISA has a Section module that is similar, but I have never used it.

Thanks all for help!

 

[Aesur]

PIEngi - I'm glad you mentioned about the location in Shapebuilder, I actually forgot about that part, so once you region the shape you can run "massprop" command in AutoCAD and find centroid and then move it to 0,0 in the CAD file and when you import it will fix the grid issue for you.

 

[wrantler]

I would agree with KootK.

Last I remember, trying to refine the Cw/J numbers did not really make much difference than If I used conserv values.. Maybe increased capacity a few %

When it comes to custom built up sections I tend to be conserv on estimating their capacities.

 

[271828]

For this, I feel that it will be reasonable and conservative to calculate J based on the four plate elements that make up the box and to ignore the warping contribution.

That's what I would do.

To dig deeper would result in a pretty hard problem. Advanced Mechanics of Materials textbooks have examples that are fairly similar.

 

[KootK]

Thanks for the votes of confidence guys. My guess for this is that OP will be applying a buckling provision that requires the torsional stiffness of the thing being designed as an input parameter. As such, I feel that the most direct path to getting to "done" is:

1) Conservatively but reasonably estimate the torsional stiffness by hand as discussed.

2) Re-enter the design standard to see if buckling is now further down the food chain than other failure modes, as I expect.

3) Get fancy with J & Cw only in the unlikley event that #2 doesn't pan out.

When it comes to buckling, usually what you need from a member's torsional stiffness is "enough to keep moving" rather than "wholly accurate".

Granted, if a designer is not comfortable with the theory underlying the approximation, then the FEM double check is effort well spent.

 

[PIEngi]

Kootk- I pretty much did what you outlined. I did end up modeling this in the python program Lomarandil linked above, mainly for experimenting between full and intermittent connectivity between the parts (which ShapeBuilder won't do) and seeing the magnitude of difference between the two. I'll also have configurations with different channel sizes, and the python module will make it easy to automate the calc printouts.

The results: the hand calcs worked out fine lol. They were much more conservative than the FEA results (FEA torsion constant was about 40% higher), but it didn't end up mattering based on the way LTB checks are setup in the AISI code. I am glad I ran through this exercise though, as I found that the torsion constant from ShapeBuilder was orders of magnitude higher than the hand calcs and the python results. Something to keep in mind for other projects where this may actually matter.

Thanks everyone for the great discussion.