Seismosoft | Earthquake Engineering Solutions

Hello,

I am modelling a building in SS (v6.5) composed of RC shear walls supporting simple slabs, and I was wondering how to model one of the walls I have. It has a monosymmetric double-T section with a web of 3.3m, a flange of 5.7m, and one of 3.9m. I was wondering if there is a module or something that takes this section shape into account or not (I've only seen it for steel profiles).
Alternatively, do you have an idea as to how to model it otherwise? I will launch all kinds of analysis (pushover, time-history, bidirectionnal, etc.). I thought about creating one section with the web and the big flange, and another one with the small flange and link the two elements rigidly (I am brand new to SS, but I assume that is feasible, is it? and how?)

Thank you very much in advance for your advices.

Interesting problem, Matcamp. Another option worth looking into might be the reinforced concrete U-shaped wall section - 'rcuws'.

I mention this possibility because the legs of the U do not have to be the same and you may, thus achieve the non-symmetry.

If it doesn't throw off torsional effects then the U-shape could mimic the double-T. I suppose the primary thing would be to get each of the 3 portions - web and 3 flanges - properly located in plan so as to preserve eccentricities between the center of stiffness and the center of mass along each of the two horizontal axes.

A third option would be to model using 3 rectangular sections. The feasibility of such a scheme would likely depend upon the height to length ratio of the various elements. It could be that the web takes virtually all of the transverse load in its plane with little participation by the webs and vice-versa.

Certainly, your proposal is a valid try too. The disadvantage being that you will have lots more elements since you would likely subdivide the web and flanges in order to be convinced that you, in fact, have achieved the rigid attachment you desire.

Best of luck Matcamp.

Hmmm, I also considered using two U-shaped sections, but I would loose the accuracy for the loading in the direction of the flanges, wouldn't I?

Also, what would be the advantage (if there are some) of using three rectangular sections over using one T-section and one additional rectangular section? (In my case I have no confinement reinforcement so that shortcoming of the T-section is not a problem for me)

Finally, about the number of elements and IP, for the other walls I could model the whole section (L, T, rect.) in section so I thought of using one force-based element per floor, with 6 integration points. I think it would allow good accuracy without having too much localization issues. But if I wanna link a T-section with a rectangular one to simulate an H-section, I should use more than one element per floor. Do you reckon using two elements wit 3 integration points would be sufficient?

That's a tough question about how many elements per floor to mimic full continuity between two elements. I would think more along the lines of 4 or 5 elements per column to get them adequately linked together. Having said that, the only way to really tell would be to do a parametric study of response versus number of elements and see if a trend can be identified.

Regarding the use of U-sections, I see your point about losing the accuracy in the plane of the flanges. I had been thinking more along the lines of a single U-section, but the same problem would exist.

And yes, you would still have to subdivide into more elements with 3 rectangles.

So your idea is probably the best with the proper number of subdivisions per floor.

Best of luck Matcamp.

Dear Matcamp,

Considering the dimensions and perpendicularity of the walls, you might only need to rigidly connect the walls at the top and at the bottom of each floor. If this does not provide enough accuracy, you might want to use 2 force-based elements with 3 integration sections each, or 4-5 displacement based elements per storey.
Regarding the section shapes, probably Huffte's suggestion of using rectangular sections is the simplest and most efficient one.

In any case, a sensitivity study is probably needed.

SeismoSoft Support

Matcamp,

I searched this Forum for "wall modelling" and found this previous post (http://www.seismosoft.com/forum/topic.asp?TOPIC_ID=92) that you may find interesting, given that it refers to a publication describing numerical vs. experimental comparisons for U-shaped walls (this publication is also cited in SeismoStruct's User Manual/Help System).

Rui

First things first: thank you all for the answers.

@Seismosoft: So you would rather advise to model three (rigidly linked) rectangular elements than one T-shaped and one rectangular? I have already prepared the sections for the latter, but changing it wouldn't be much trouble. I think I might try both solutions and see what comes out.

@ruipinho: Thanks, it is actually my supervising professor's PhD thesis, and I already had to read it (at least the parts that were relevant to my thesis).

Last question on this topic: To link them, I should use "Rigid Link", right? not "Equal DOF"?

Correct.

But you could also have experimented things a bit before, e.g. by trying an elastic very stiff beam, or a beam with rigid offsets, or 'rigid links' constraints, or 'equal DOF' constraints, then comparing it all and withdrawing your own conclusions.

Rui