Accidental Torsion in Seismic Analysis

[rickfischer51]

I am performing a seismic analysis on an apparatus per ASCE-4. I am using an equivalent static load per the code and applying it to a full nonlinear finite element model. The model is fully detailed (distributed mass, no stick and lumped mass idealizations, etc). I'm now being told I need to account for accidental torsion. My interpretation after reading the code and the commentary is that a detailed finite element model accounts for this (3.1.1.(d) The model shall represent the actual locations of the centers of masses and centers of rigidity, thus accounting for the torsional effects caused by the eccentricity) and that this is only required when an idealized model is used that is "apparently symmetric." Do I really need to account for accidental torsion?

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[KootK]

I believe that you do still need to account for accidental torsion.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[rickfischer51]

OK, but why? I'm trying to understand the physical significance of this.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[KootK]

Torsion can have some pretty deleterious consequences if not accounted for. For conventional structures, it's rather difficult to pin down either the centre of rigidity or the centre of mass with much accuracy. If your structure is symmetric, the accidental torsion is a must. If it's heavily asymmetric, than the accidental torsion probably won't be much of a penalty anyhow.

What exactly is an "apparatus" anyhow?

There is code committee talk afoot that may lead to elimination of accidental torsion requirements being waived for structures analysed with non-linear algorithms (RSA/THA).

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[structSU10]

As I understood it, accidental torsion is due to whatever is being built not being build as ideally as we model it. Mass isn't as evenly distributed in reality as in our models, so the 5% eccentricity is to account for that, along with stuff that may not have been accounted for in the model 100% accurately.

 

[rickfischer51]

KootK. Its a thick wall steel box on some tubular legs. The box is slabs bolted together. The legs are weldments that are bolted to the box and the floor. It is not symmetric. The FE model is made directly from the CAD assembly file.

structSU10: I got the impression that the problem was the use of idealized models that may use evenly distributed mass to model something that can actually vary, like a poured concrete floor, or contents of a building that are somewhat random and can get moved around. Or, lumped masses in a non-FE model. The commentary to the code says, at C3.1.1 (a)., "The amount of detail used to represent a structure in a mathematical depends on the structural configuration and the use of the model. Finite element mathematical models are used to represent complex structures..... Specific considerations and requirements pertinent to idealizing complex structures are provided here." Then at (d), "For Lumped-mass stick models...", and at (e), In an apparently symmetric system...", i.e. a lumped mass stick model. Also, in "Evaluation of Code Accidental-Torsion Provisions from Building Records" by Chopra, et al, "The additional+/-0.05b, known as accidental eccentricity, is introduced to account for building torsion arising from discrepancies between the mass stiffness and stress distributions used in the analysis and true distributions at the time of an earthquake." My finite element model is complete and detailed and unless i'm getting serious variations in the density of a rolled steel plate, I dont believe I have any such discrepancies.

I'm not trying to be argumentative here. If I gotta do, I'll do it, I just want to understand it.

Rick Fischer
Principal Engineer
Argonne National Laboratory

 

[SAIL3]

one might be able to get the center of mass with a good degree of accuracy but the center of rigidity may be less reliable...no such thing as a truly pinned or fixed connection..as-built condition can vary from assumed design condition, etc