Running problem of sfsi model in webinar25

[[email protected]]

Dear STKO Team
I try to simulate your operations related to sfsi files in webinar25, but when I calculate, OPENSEES prompts me "warning analysismode:getgroupgraph-0vertices, has the domain been populated?" Would you please have a look at my model, thank you.

[STKO Team]

You should partition your mesh before running the model in parallel with OpenSeesMP

[[email protected]]

Thank you for your reply. It is now ready to run.

[STKO Team]

You're welcome

[Charon]

You're welcome

Dear STKO team, I encountered the same issue when modeling based on Webinar24/25. My workstation has two 26-core CPUs with a total of 52 cores. Specifically, OpenSeesMP can run normally when the number of partitions is set to 26 or less. However, once the number of partitions exceeds 26, I encounter the same problem as the original post author. But I remember that my previous model could be divided into 48 parts and called to run on 48 cores without any problems. So I wonder what happened?
Additionally, my model runs particularly slow during seismic analysis in the last 30 seconds. Especially after the seismic waves started shaking during the 5s.Could you help me identify where the problem lies? I have attached my model below.
I'm not sure if you prefer discussing new issues in old posts like this; if not, next time I will create a new post for inquiries! Thank you very much!

Best wishes，
Charon

[STKO Team]

You should first try to run your model with elastic materials to see if everything is correct.
Can you do that?
In this way, we can understand if it takes time because it's failing to converge due to nonlinearities or due to some instabilities.

Also, do not use too many processors. Your model is pretty small, 8 to 12 cores will be enough.
Otherwise it will take more time (due to parallel overhead)

[Charon]

You should first try to run your model with elastic materials to see if everything is correct.
Can you do that?
In this way, we can understand if it takes time because it's failing to converge due to nonlinearities or due to some instabilities.

Also, do not use too many processors. Your model is pretty small, 8 to 12 cores will be enough.
Otherwise it will take more time (due to parallel overhead)

Dear STKO team, thank you for your help! For my previous version of the model, I made some changes and added 2% Rayleigh damping, I have attached the latest model below, and the analysis has worked fine and at a pretty good speed in the Model1 file.
However, due to an oversight on my part, the upper clay (L1 ) and the lower sand (L2 ) in the Model1 version are using the same BrickUP Element, which results in a fairly large permeability coefficient for L1 throughout the analysis (up to 100!).
After noticing this error, I modified it in Model2 by dividing the BrickUPelement into two, L1 and L2. In the L1 Element, I set the permeability coefficient to be quite small (1e-9), and it all sounds perfect, right? However, when running the analysis, after running Excavation Analysis, the pop-up "WARNING: MaterialStageParameter::setDomain() - no effect with material tag 3
WARNING: MaterialStageParameter::setDomain() - no effect with material tag 4", I don't understand why this warning happens and the analysis seems to stop after this (although the background shows that OpenSeesMP is still trying to calculate it).
The only changes I made were:
1. copied a BrickUP Element, renamed it to Soil L1 Element and changed the permeability factor to 1e-9
2. selected the InitialStateAnalysisWrapperL1 (PIMY) in the Physical Element and assigned the Soil L1 Element to it
Can you guys help me to check what is going on? Thanks a lot!

[STKO Team]

Try this:

model2.zip

(561.07 KiB) Downloaded 216 times

it's running also in the last dynamic analysis.
But it has serious convergence issues. I think mainly due to the plasticity in soil.
In fact, if you look at the 3rd stage (just after you activate plastic behavior, before adding the structure) you will see high accelerations in the areas near the holes.

Probably because your pseudo-static analyses have a too short duration (1 second). Make the duration larger to avoid inertia effects in those pseudo-static stages.

Anyway, what I changed was the maxDim in the KrylovNewton set to 2*maxiter, so that you compute the tangent matrix only once per step

[Charon]

Try this:
model2.zip

it's running also in the last dynamic analysis.
But it has serious convergence issues. I think mainly due to the plasticity in soil.
In fact, if you look at the 3rd stage (just after you activate plastic behavior, before adding the structure) you will see high accelerations in the areas near the holes.

Probably because your pseudo-static analyses have a too short duration (1 second). Make the duration larger to avoid inertia effects in those pseudo-static stages.

Anyway, what I changed was the maxDim in the KrylovNewton set to 2*maxiter, so that you compute the tangent matrix only once per step

Dear STKO Team，
Thank you for your reply and for doing a thorough check of my model! I really appreciate it, a few weeks have passed and I have gone through the posts on the forum and made changes to my model with your suggestions and now the model runs perfectly to the end and runs fast. However, he still has a couple of minor problems, which I would like to present to you:
1. I found when searching the forum for BrickUP units that due to the special nature of UP series units, it seems we cannot apply self weight to the soil by applying VolumeForce through a LoadPattern, but only by changing bZ to -9.81 in the Element Property of the UP unit to apply self weight. Am I understanding this correctly? This is exactly how my current model is running.

2. In some of my analysis commands such as Excavation Analysis, Settlement Analysis, and plastic analysis I use Constraints as Plain or Transformation instead of the common Penalty because I found that Penalty will run without convergence, while Plain or Transformation is fast and convergent. Do you think this is feasible in my model?

3. Is there any problem with my analysis step settings? Do you have any suggestions? For example, in terms of boundary conditions or interactions, or the order of analysis steps?
In my model, I have used InitialStateAnalysisON and InitialStateAnalysisOff twice, which I found when I looked up the forum that others have done this, and I don't know if I should do this, or if I should start InitialStateAnalysisON at the beginning and finish running the plastic analysis and then InitialStateAnalysisOff?
To be honest, this has been bothering me.

4. The acceleration results of my analysis runs show a weird trend（As shown in the picture）. The acceleration at the deepest part of the soil is quite small, and the acceleration at the shallower part gets larger, which I know should be the rule, but the acceleration peaks at the deepest and shallowest parts are off by several orders of magnitude! In my perception, the deepest acceleration should be approximately equal to the peak of the UniformExcitation input (0.2) Can you see how this works? Or is there something wrong with the way I input seismic waves?

The curve of the soil pore water pressure is also strange, the pore water pressure does not rise with vibration, and I can't figure out why. This really bothers me.

5. As a newbie learning finite element modeling, I am eager to see when there will be a book to guide STKO modeling in all directions (e.g., structural, geotechnical). Do you have any news on this?

Best wishes，
Charon

[STKO Team]

1. I found when searching the forum for BrickUP units that due to the special nature of UP series units, it seems we cannot apply self-weight to the soil by applying VolumeForce through a LoadPattern, but only by changing bZ to -9.81 in the Element Property of the UP unit to apply self-weight. Am I understanding this correctly? This is exactly how my current model is running.

No. You can apply self-weight using Volume Force (it works for any kind of 3D element). What we said in other posts is that when you want to apply body force from the element definition (and not from Volume force) you have to pay attention to this:
With standard (displacement-based) elements, you input the body force itself (-g * rho), while with UP elements you input only -g.

2. In some of my analysis commands such as Excavation Analysis, Settlement Analysis, and plastic analysis I use Constraints as Plain or Transformation instead of the common Penalty because I found that Penalty will run without convergence, while Plain or Transformation is fast and convergent. Do you think this is feasible in my model?

It depends on the kind of boundary conditions you have.

3. Is there any problem with my analysis step settings? Do you have any suggestions? For example, in terms of boundary conditions or interactions, or the order of analysis steps?

We cannot tell you if it is fine or not, it depends on what you want to do

In my model, I have used InitialStateAnalysisON and InitialStateAnalysisOff twice, which I found when I looked up the forum that others have done this, and I don't know if I should do this, or if I should start InitialStateAnalysisON at the beginning and finish running the plastic analysis and then InitialStateAnalysisOff?
To be honest, this has been bothering me.

The InitialStateAnalysisON/OFF pair should be used only ONCE!

4. The acceleration results of my analysis runs show a weird trend（As shown in the picture）. The acceleration at the deepest part of the soil is quite small, and the acceleration at the shallower part gets larger, which I know should be the rule, but the acceleration peaks at the deepest and shallowest parts are off by several orders of magnitude! In my perception, the deepest acceleration should be approximately equal to the peak of the UniformExcitation input (0.2) Can you see how this works? Or is there something wrong with the way I input seismic waves?

You are using Uniform Excitation so it make sense. The output will be relative acceleration(relative to the fixed nodes at the bottom level)

5. As a newbie learning finite element modeling, I am eager to see when there will be a book to guide STKO modeling in all directions (e.g., structural, geotechnical). Do you have any news on this?

Little by little we are building our e-learning courses... but it takes time