Soil Structure Interface Modeling

[kolozvari]

Thank you so much for helping with the geometries. Can you please remove the model from forum?

I have a few follow up questions:
1) In the model you sent, it seems that all coupling beam rigid links and rigid diaphragms are deleted at most floors except 2 levels. Why is this? Do I need to re-assign them?
2)

The foundation podium should be cut as well

Why is this necessary? Foundation and basement and core walls were slaved with equalDOF along the edges. Their geometries were not compatible in the previous version of the model and it worked fine. Please clarify.
3)

The slabs must be cut as well, because your connection beams where split in 2. The node in the middle will cut the slab faces at that point, so you need to extend a cut also there. After that re-assign mesh-controls

I actually purposly left this disconnected because I plan to have a shear hinge in the middle of the beam in the nonlinear model and I dont want a slab to interact. Is there an easy way to disconnect them in the present model?
4)

Final Check!!!! IMPORTANT

. Are you saying that I need to re-assign properties of coupling beams? That's not a problem.

Looking forward to hearing from you.

Thank you.

[STKO Team]

1) In the model you sent, it seems that all coupling beam rigid links and rigid diaphragms are deleted at most floors except 2 levels. Why is this? Do I need to re-assign them?

Yes, it's the same issue as per the physical & element property assignment.
When you merge 2 geometries with coinciding entities (edge in this case), some property/condition can be removed....

2)
The foundation podium should be cut as well
Why is this necessary? Foundation and basement and core walls were slaved with equalDOF along the edges. Their geometries were not compatible in the previous version of the model and it worked fine. Please clarify.

It's not mandatory, but it will be more accurate. Because you are attaching them with EqualDOF, so it is consistent if the nodes share the same location (no distance between slave and master). And this cannot be guaranteed if the mesh are incompatible

3)
The slabs must be cut as well, because your connection beams where split in 2. The node in the middle will cut the slab faces at that point, so you need to extend a cut also there. After that re-assign mesh-controls
I actually purposly left this disconnected because I plan to have a shear hinge in the middle of the beam in the nonlinear model and I dont want a slab to interact. Is there an easy way to disconnect them in the present model?

Oh I see. So in this case you should have done it in a slightly different way. you can keep the slab detached from the main structure. Then you can attach it back with equalDOF, only using the edges that you want to actually link (in this case you should skip the beams).

4)
Final Check!!!! IMPORTANT
. Are you saying that I need to re-assign properties of coupling beams? That's not a problem.

Yes

[kolozvari]

Thank you for your responses.

Oh I see. So in this case you should have done it in a slightly different way. you can keep the slab detached from the main structure. Then you can attach it back with equalDOF, only using the edges that you want to actually link (in this case you should skip the beams).

I dont really understand what you mean by this. However, this is not an issue at the moment and I might contact you again regarding this later when I start working on the nonlinear model.

Also, I have a couple of questions regarding gravity load application. In my previous models all gravity load was applied via body forces and I didnt have any load patterns defined in the model and in the analysis steps. However, in my latest model I defined area load on each floor via FaceForce, which should be applies in addition to the body force. I assigned this load to loadPattern command in the Analysis Steps. In my understanding, when I run gravity analysis, both body forces and forces applied to this load pattern will be both considered. Can you please confirm that my understanding is correct?

When I run the model I am getting this error.

TimeSeries *getTimeSeries(int tag) - none found with tag: 0
WARNING - problem creating TimeSeries for LoadPattern 28

while executing
"pattern Plain 28 0 {

# Loads.Force FaceForce

if {$STKO_VAR_process_id == 0} {
load 61421 0.0 0.0 -2.797129999999996 0.0 0.0 0.0
load 42380 0.0 0.0..."
(file "analysis_steps.tcl" line 848)
invoked from within
"source analysis_steps.tcl"
(file ".\main.tcl" line 79)
Process Terminating 0
Using ASDShellQ4 - Developed by: Massimo Petracca, Guido Camata, ASDEA Software Technology
TimeSeries *getTimeSeries(int tag) - none found with tag: 0
WARNING - problem creating TimeSeries for LoadPattern 28

while executing
"pattern Plain 28 0 {

# Loads.Force FaceForce

if {$STKO_VAR_process_id == 0} {
load 61421 0.0 0.0 -2.797129999999996 0.0 0.0 0.0
load 42380 0.0 0.0..."
(file "analysis_steps.tcl" line 848)
invoked from within
"source analysis_steps.tcl"
(file ".\main.tcl" line 79)
Process Terminating 1

When I look at my load pattern definition, I can see that tsTag = 0, which seems that causes the error. Do I need to defined some king of a time series to apply the load distributed on the slab? How does this combine with gravity load from body forces? Can you please clarify?

Thank you!

[STKO Team]

I dont really understand what you mean by this. However, this is not an issue at the moment and I might contact you again regarding this later when I start working on the nonlinear model.

I mean that you can use the same strategy I used for connecting the walls (shell) and the podium (solid).
You keep 2 geometries detached (main structure and slabs), then you create an interaction between the slab's edges and the main structure's walls and columns (excluding the connecting beams). This interaction will have an equalDOF in all 6 DOFs. By doing so, the slabs and the structure will be effectively "linked" everywhere but in the connecting beams.

When I run the model I am getting this error.
TimeSeries *getTimeSeries(int tag) - none found with tag: 0

Yes, you need a timeSeries: OpenSees does not allow a loadPattern without a timeSeries. A timeSeries is what tells the load how should evolve in time.

The bodyForce only works with continuum (solid 3D or 2D) elements. It's directly input in the definition of the solid element, and it computes the vector of external forces given by the bodyForces when computing the unbalance vector.
Note: the bodyForce in the element definition is ALWAYS applied instantaneously (because it is not associated with a timeSeries, where you can instead use a linear timeSeries to make load grow linearly)

So you have 2 options:

1. BodyForce(in solid element property) + FaceForce (in condition on faces, in a loadPattern before the gravity analysis)
   The BodyForce will be applied instantanously, while the FaceForce will follow the timeSeries you use in the loadPattern. So, if you want to be consistent, the loadPatterm should have a constant time as well.
   Problem: both loads will be applies instantaneously. If they are large and trigger non-linear behaviors, it can give convergence issues that cannot be solved using adaptive time steps (because the timeSeries is constant!)
2. VolumeForce (in condition on solids) and FaceForce (in condition on faces), both in a loadPattern. In this way you can assign a linear TimeSeries and both will evolve incrementally (ideal for nonlinear analyses)

The two approaches are theoretically equivalent. VolumeForce is an STKO automation that takes a force-per-unit-volume from the users, and converts it into equivalent nodal forces using the shape functions of the solid element. Exactly as OpenSees's solid elements do with their own body forces

[kolozvari]

Thank you for this explanation, it is very useful. I decided to go with option 1 for now (it was easier to implement). The model seems to be running fine. Once I start running nonlinear analyses I might use option 2 to avoid any convergence problems.

Thank you very much!

[STKO Team]

You're welcome

[kolozvari]

Hello! I continue to refine my model and I have added slabs (no more rigidi diaphragms) and columns, distributed mass (used to be lumped), and gravity load. The model with SSI and full soil domain runs smoothly, but the fixed base model is not converging (again) for dynamic analysis (gravity runs OK). If you remember, we had similar issue a couple of weeks ago, where using ASDshell fixed the problem. In this case, however, this doesn't seem to work. I am not exactly sure how to tackle this issue. Could you please take a look at the model (attached) and let me know what you think?

Thank you!

[STKO Team]

Look at these warnings when you run the analysis:

ParallelSettings.png (18.38 KiB) Viewed 192 times

In the Gravity analysis you did not use the proper settings for parallel analysis (numberer = RCM and system = UmfPack). Doing so, each partition will be detached from each other. It's mandatory to use Parallel RCM and MUMPS

[kolozvari]

Ah, my bad. I am sorry, should have known better by now. I used the same model for modal analysis and forgot to change this.

Thank you!

[STKO Team]

You're welcome