Frame with beam span loads

[akshay somkuwar]

Hi
I am conducting a simple frame analysis using a model derived from the verification example in the ETABS manual. Most of the calculated bending moment values exhibit an error of less than 5%. However, I have observed a significant deviation at one location in Load Case 1, where the error exceeds 20% (highlighted in red in the attached table). Additionally, for Load Case 2, the error reaches 100% (highlighted in red).

To better understand the cause of these discrepancies, could you please clarify:
1. Which integration method is used beam elements in NextFEM(e.g. Gauss-Legendre with 5 integration points or Gauss-Lobatto with 5 integration points)?
2. How is the extrapolation of values performed in NextFEM for computing bending moments?

For your reference, I have attached the model file and the analysis PDF.

Thank you for your time and assistance.

Attachments:

You must be logged in to view attached files.

[NextFEM Admin]

Hello,
the differences you registered are not related to integration method or extrapolation method, but more generally to the beam formulation the solver uses. In that sense, the result are solver-dependent, and they should be referred to the built-in solver rather than to NextFEM.
In particular, you’re showing the results form OOFEM (the default solver); please refer to their website to get more info on beam formulation and how results are recovered. You can try running the model with OpenSees directly in NextFEM Designer, by changing Solver options.
Finally, by our experience, linear analyses with beams can differ from one solver to another: OOFEM shows little difference with CSI product solvers, while it has good agreement with others. Don’t expect the same stiffness matrix for the same element in all solvers.
In addition, some statistical remarks: you cannot refer the COV (Coefficient of Variation) on null values: your last red row shows a difference that, in absolute terms, is comparable to the other lines.

[akshay somkuwar]

Thank you very much for the clarification and your remarks.
My understanding was that the linear response should not change between different solvers.
Although the OOFEM website mentions that the beam2d element is based on the Timoshenko hypothesis, disabling the “include shear deformation in element” option essentially converts the formulation to the Euler-Bernoulli hypothesis, similar to how it is implemented in CSI software.

I will try running the model in OpenSees.

[NextFEM Admin]

CSI software also uses Timoshenko hypothesis, unless you can disable it. In any case, there are many things that can cause differences (e.g. auto-meshing in CSI, etc.).

[Author]

Posts