asdeasoft.net

Hi STKO Team!

I want to ask about a convergency issue that I am dealing in a confined masonry structure subjected to acceleration recording in the X and Y directions. In the document CM Building that I attach you can find this STKO model with the planes of the structure which were the used information to model. I used ASDConcrete3D for the masonry panel, and the RC columns and beams, while ElasticMembranePlateSection was used for the slabs. The element used was ASDShellQ4. For the tensile and compressive fracture energies of the masonry panel I used Gc = 15 + 0.43 fc - 0.0036 fc^2 and Gt = 0.025 (2 ft)^0.7, which were the expressions that worked better in a previous calibration process that I carried out in a set of 2D walls subjeted to cyclic displacement. The registers used are attached below in the image called Registers. Previously, a modal analysis of the structure was carried out, with the results attached in the Modal.zip document.

The main problem that I am dealing with in the seismic analysis is the non-convergency of the model due to the concentrations of deformations in some parts of the structure as you can see in the image called Non-Convergency that I had attached. I tried to solve that by generating a TCL script which use differents algorithms, but I couldn't pass this convergency issue.

I wanted to know if you can check my model and see if something related with the modelling or the choosen materials is generating the issue. I was thinking that the issue is a numerical one, but simply it could be maybe the collapse of these walls as well.

Many thanks in advance!

Hi again!

Before your response, I want to make an update below: I have been trying a lot of changes in the model like the use of EnergyIncr as convergency criteria, adding viscosity in the ASDConcrete3D materials, using the TRBDF2 integrator instead Newmark 0.5 0.25, among others. With this changes, I improved a little bit my results, but I can't reach a reliable convergency during all the time of the registers, especially because there is a moment in which the unbalance increases too much (1e+08 or something like that); after that I could advance but only using the ExpressNewton algorithm, which is an algorithm that always advance no matter the value of the residual displacement/unbalance. I can imagine that I have a situation with the high frequency modes that could be solved by the TRBDF2 algorithm by adding numerical damping.

Regarding all this, I want to make some questions:

1) Is there any theorical conflict if I use TRBDF2 with Krylov-Newton with maxDim > 1?

2) Is there any problem if I use a viscosity value in ASDConcrete3D? Which is a good relation between this number (which is a relaxation time) and the time step of the registers?

2) Do you think that this convergency problems are related with the geometrical features of the structure? I tried with another pair of registers with a bit less accelerations's levels and I could reach full convergency, so maybe the registers that I first sent are too much for the model in terms of PGA and frequency content. I suggest that because I really have been trying a lot of possible improvements during this week, and nothing completely works with this particular registers.

Many thanks in advance, and sorry for the long commentators but this is really important for me!

Try this: I think you forgot vertical loads

Oh, I thought that adding a mass density in the ASDConcrete3D materiales would be enough. Always I need to add the FaceForce aswell?

Masses in OpenSees are not directly converted to vertical loads (because you never add vertical gravity acceleration)
So you need to apply vertical loads manually in a static (or slow dynamic) analysis

But in a dynamic analysis like the one I am performing, do i need to add mass density + vertical loads? or with mass density is enough?

Mass generates inertia forces due to acceleration. But typically, in a FEM model you only apply horizontal (and sometimes vertical) acceleration due to seismic actions.
You almost never input a constant vertical acceleration equal to the acceleration of gravity.
The typical workflow (not only in STKO and OpenSees) is to apply mass (for dynamic analysis) and the corresponding self weight (for the static vertical analysis)