Hello there,
I want to do a 3D Site response analysis in which elastic soil properties vary three dimensionally according to geotechnical and geophysical data retrieved on surveys. I have two questions:
1) Is it possible to generate my 3D Model (i.e G(x,y,z)) and map in onto STKO?, In a webinar (https://www.youtube.com/watch?v=TrHhvJ8L9ao&t, min 15:00 approx) you create a random gaussian soil model and map it into STKO using the Python API, but I have searched in the webinars and you don't elaborate further about this. I want to do something similar, but my data is not random but rather deterministic.
2) If I have varying soil properties for each hexahedron in my model, should I match the lateral absorbing boundaries properties to each hexahedron or this would be an overkill?
If generating the 3D variable model is not possible within STKO, I was thinking in creating the mesh in STKO, and then customize the .tcl script with a Python script and assign different material properties, but this might be cumbersome. Do you have any suggestions?
KInd regards,
Sebastián
Map 3D Soil Variabilities into STKO soil model
Re: Map 3D Soil Variabilities into STKO soil model
You can do it quite easily. We just developed a tool for this. But it's not yet documented so I will try to explain it here.
The tool is the RandomizedMaterialWrapper. It is an STKO's wrapper that allows you to select a source nD material, and a text file that represents a Random Material Table. STKO will create modified versions of the source nD material just before running the analysis.
Try to follow these instructions and let us know if something is not clear.
In the attached file you will have 2 regions (one for absorbing boundaries, and one for soil elements) and 2 custom scripts required to make the update work correctly. The RandomizedMaterialWrapper only affects the material properties of soil. But the absorbing element should match the corresponding soil elements. This is exactly what those scripts do.
After them, I just put a "record" command to validate the procedure. You can go on with your analysis steps.
Assume you have 9 sample points from geotechnical surveys...
Define a RandomizedMaterialWrapper for soil elements...
Define the Random Material Table file...
This is what the results should look like
The tool is the RandomizedMaterialWrapper. It is an STKO's wrapper that allows you to select a source nD material, and a text file that represents a Random Material Table. STKO will create modified versions of the source nD material just before running the analysis.
Try to follow these instructions and let us know if something is not clear.
In the attached file you will have 2 regions (one for absorbing boundaries, and one for soil elements) and 2 custom scripts required to make the update work correctly. The RandomizedMaterialWrapper only affects the material properties of soil. But the absorbing element should match the corresponding soil elements. This is exactly what those scripts do.
After them, I just put a "record" command to validate the procedure. You can go on with your analysis steps.
Assume you have 9 sample points from geotechnical surveys...
- RMAT_1.png (112.27 KiB) Viewed 3816 times
Define a RandomizedMaterialWrapper for soil elements...
- RMAT_2.png (29.1 KiB) Viewed 3816 times
Define the Random Material Table file...
- RMAT_3.png (33.37 KiB) Viewed 3816 times
This is what the results should look like
- RMAT_4.png (79.81 KiB) Viewed 3816 times
Re: Map 3D Soil Variabilities into STKO soil model
Hello STKO Team,
Thank you very much for your thorough answer. I took a while in answering because I wanted to wrap-up all my arising questions.
I) Here is a 3D Model (base_3dmodel.scd) I built with the RandomMaterialWrapper, it looks quite nice. As you can see however, the corner absorbing boundaries weren't updated. This is because I couldn't find a way to include them in the Absorbing Boundary selection set with the STKO GUI (If you see the elements.tcl and analysis_steps.tcl, you will notice that the tags of the corner nodes aren't included). I don't know if there is a way to include them with the GUI, but if it's not possible I can do it by scripting or even neglect them.
II) I wanted to know how the randomization is performed. From the geotechnical survey data I have, the 3D Model parameters were interpolated with krigging, so in a way, there is no need to randomize further. Is there a way to set the variance from the RandomMaterialWrapper to zero? Again, if it is not possible, the effect it's perhaps negligible.
III) The soil model I randomized its a parallepiped so the meshing was straightforward. The top of the model however, is a hill with a lot of irregularities, so I'm having trouble in drawing and meshing layered_topography.scd:
I did the drawing by layers using the Python API and STKO, but it was kind of grueling and prone to errors, and I don't have experience with CAD programs at all. Also, when I try to mesh, even if I have only planar faces, it doesn't create brick elements, what I'm doing is correct or should I find another approach?
Thank you!
Thank you very much for your thorough answer. I took a while in answering because I wanted to wrap-up all my arising questions.
I) Here is a 3D Model (base_3dmodel.scd) I built with the RandomMaterialWrapper, it looks quite nice. As you can see however, the corner absorbing boundaries weren't updated. This is because I couldn't find a way to include them in the Absorbing Boundary selection set with the STKO GUI (If you see the elements.tcl and analysis_steps.tcl, you will notice that the tags of the corner nodes aren't included). I don't know if there is a way to include them with the GUI, but if it's not possible I can do it by scripting or even neglect them.
- random_model.png (388.92 KiB) Viewed 3783 times
III) The soil model I randomized its a parallepiped so the meshing was straightforward. The top of the model however, is a hill with a lot of irregularities, so I'm having trouble in drawing and meshing layered_topography.scd:
- topography.png (209.18 KiB) Viewed 3783 times
Thank you!
- Attachments
-
- 3DRandomModel.rar
- (896.33 KiB) Downloaded 205 times
Re: Map 3D Soil Variabilities into STKO soil model
You are right, the script I sent you works for the 2D case.
Here's a more generic script. It's even easier because you don't need to create sets: absorbing and soil elements will be recognized automatically from their class tag from OpenSees. Regarding your second problem:
It will not mesh it with Hexa elements because you need to build a geometry made of 6-face solids so that it will be suitable for structured meshing.
Then, you also need to make a single geometry otherwise each layer will be detached from the other.
Unfortunately for this kind of model you need to gain a bit of experience with CAD modeling.
You should start by looking at the webinars about geometry modeling from the "get started in opensees with stko" webinars.
Hint:
You should draw smooth countour lines for each level (with splines) and join them with the Loft command. It will create a smooth surface. Then you extrude it to make a solid
Here's a more generic script. It's even easier because you don't need to create sets: absorbing and soil elements will be recognized automatically from their class tag from OpenSees. Regarding your second problem:
It will not mesh it with Hexa elements because you need to build a geometry made of 6-face solids so that it will be suitable for structured meshing.
Then, you also need to make a single geometry otherwise each layer will be detached from the other.
Unfortunately for this kind of model you need to gain a bit of experience with CAD modeling.
You should start by looking at the webinars about geometry modeling from the "get started in opensees with stko" webinars.
Hint:
You should draw smooth countour lines for each level (with splines) and join them with the Loft command. It will create a smooth surface. Then you extrude it to make a solid
Re: Map 3D Soil Variabilities into STKO soil model
This is awesome! Thank you very much.
Now I have to work on the mesh, I will update when I have a result!
Now I have to work on the mesh, I will update when I have a result!
Re: Map 3D Soil Variabilities into STKO soil model
Hello, I found a problem earlier than I thought!
I created smooth splines for each level as you said.
When I join them with the loft command the result gets messy and the volume can not be meshed (I always use the ruled loft algorithm):
I guess this is because layers are too different from each other, so I worked on some solutions. Here I created two different lofts at the top and the bottom. The hexameshing worked fine for each loft separately, but I could not find a way to join them. Moreover, some layers at the center were too problematic to loft, so I would have to ignore them and change the real topography.
I'm pretty sure the problem could be solved if I cut the lower layers and form triangles (see the green lines in the picture behind) instead of irregular pentagons. Again the problem here would be that I'm ignoring an important part of the topography.
Do you have any suggestions? Should I just simplify the topography? I attach the model with the spline wires.
I created smooth splines for each level as you said.
- smooth_splines.png (43.2 KiB) Viewed 3732 times
- problem_lofts.png (96.34 KiB) Viewed 3732 times
- potential_solution_1.png (50.5 KiB) Viewed 3732 times
I'm pretty sure the problem could be solved if I cut the lower layers and form triangles (see the green lines in the picture behind) instead of irregular pentagons. Again the problem here would be that I'm ignoring an important part of the topography.
- potential_solution_2.png (42.16 KiB) Viewed 3732 times
- Attachments
-
- layered_model_splines.rar
- (15.2 KiB) Downloaded 180 times
Re: Map 3D Soil Variabilities into STKO soil model
The curve levels should not be done in this way. Some of them cross each other. Others are overlapping at the same Z Coordinate.
This makes the Loft process non-unique.
However, we are adding a new feature that will be released soon. It allows you to create a B-Spline Surface (free-form) from points or curves used as constraints.
In the meanwhile, I used our tool (work in progress) on your model. I had to clean some of your iso-curves to make them consistent (no crossing and no overlap).
Here I obtained a surface that passes through your curves.
Then I extruded it to make a volume.
Let us know
This makes the Loft process non-unique.
However, we are adding a new feature that will be released soon. It allows you to create a B-Spline Surface (free-form) from points or curves used as constraints.
In the meanwhile, I used our tool (work in progress) on your model. I had to clean some of your iso-curves to make them consistent (no crossing and no overlap).
Here I obtained a surface that passes through your curves.
Then I extruded it to make a volume.
Let us know
- Attachments
-
- layered_model_splines.zip
- (63.39 KiB) Downloaded 176 times
Re: Map 3D Soil Variabilities into STKO soil model
Hello,
Again, thank you very much for the explanation and the building of the geometry. Looking forward for the implementation of all those fantastic tools you are developing.
The analyses went super smooth and the effects of 3D elastic soil properties in the seismic performance are clear. I have just a few questions.
1) Is there a problem if I run the analyses on MP? I did some runs on a sequential and parallel already and the results were the same, so I just want to confirm.
2) When I open the script editor and load a .py file, I get a ScriptDocument error, then the file cannot be read or edited in the editor. I can edit it with a text editor outside, and it has no issues with running. Not sure why this is happening.
3) I'm most interested in the surface nodal accelerations and velocities. Since selecting the nodes in the surface plot is kind of difficult from the render view when there is topography, I created a selection set with the surface and recorded the output and retrieved the chart data. Is there a better approach for achieving this?. I think it may be possible to do it with the pympc getNodalResult function (you did something similar in the introduction to python API webinar, but I got lost on how to filter and select from nodal coordinates).
Kind regards,
Sebastián.
Again, thank you very much for the explanation and the building of the geometry. Looking forward for the implementation of all those fantastic tools you are developing.
The analyses went super smooth and the effects of 3D elastic soil properties in the seismic performance are clear. I have just a few questions.
1) Is there a problem if I run the analyses on MP? I did some runs on a sequential and parallel already and the results were the same, so I just want to confirm.
2) When I open the script editor and load a .py file, I get a ScriptDocument error, then the file cannot be read or edited in the editor. I can edit it with a text editor outside, and it has no issues with running. Not sure why this is happening.
3) I'm most interested in the surface nodal accelerations and velocities. Since selecting the nodes in the surface plot is kind of difficult from the render view when there is topography, I created a selection set with the surface and recorded the output and retrieved the chart data. Is there a better approach for achieving this?. I think it may be possible to do it with the pympc getNodalResult function (you did something similar in the introduction to python API webinar, but I got lost on how to filter and select from nodal coordinates).
Kind regards,
Sebastián.
Re: Map 3D Soil Variabilities into STKO soil model
There should be no problem, it should work just fine1) Is there a problem if I run the analyses on MP? I did some runs on a sequential and parallel already and the results were the same, so I just want to confirm.
Can you show a screenshot of the error?2) When I open the script editor and load a .py file, I get a ScriptDocument error, then the file cannot be read or edited in the editor. I can edit it with a text editor outside, and it has no issues with running. Not sure why this is happening.
You can for example create a script that searches for the node nearest to a user-defined X-Y-Z coordinate.3) I'm most interested in the surface nodal accelerations and velocities. Since selecting the nodes in the surface plot is kind of difficult from the render view when there is topography, I created a selection set with the surface and recorded the output and retrieved the chart data. Is there a better approach for achieving this?. I think it may be possible to do it with the pympc getNodalResult function (you did something similar in the introduction to python API webinar, but I got lost on how to filter and select from nodal coordinates).
Try to do it and post here any doubt
Re: Map 3D Soil Variabilities into STKO soil model
Dear STKO Team,
I found the surface nodes with a little bit of cheating. If I define a region in OpenSees I can use the NodeRange info at the analysis_steps.tcl to get the surface nodes.
Then I import a .txt with the nodetags and I do the same you did with the drift recorders in the Introduction to Python API webinar. However, I have an issue with this approach:
I) The way I call db.getNodalResult.evaluate(timestep) for each time step will store a database with more than ten thousand nodes for my model, for each time step (4000 time steps). This is extremely slow. Is there a way to trim the database in a way it considers only a subset of the total database? I'm thinking in something like this:
DB= db.getNodalResult("Displacement", match = MpcOdb.Contains)
DB_subset=DB.getnodes(if id.node == match)
and then iterate for all time steps...
I think in the database methods there are some boolean options (you did something like this in the critical damage virtual result from the same webinar), but I could not figure how to do it myself for my case.
As for the problem with the scripts not showing, I think it is something with OneDrive. If I store the scripts in a not-cloud folder this problem doesn't appear. Also when I record an animation I have to store it out of the cloud. I'm not sure why this happens, but it's ok.
I found the surface nodes with a little bit of cheating. If I define a region in OpenSees I can use the NodeRange info at the analysis_steps.tcl to get the surface nodes.
Then I import a .txt with the nodetags and I do the same you did with the drift recorders in the Introduction to Python API webinar. However, I have an issue with this approach:
I) The way I call db.getNodalResult.evaluate(timestep) for each time step will store a database with more than ten thousand nodes for my model, for each time step (4000 time steps). This is extremely slow. Is there a way to trim the database in a way it considers only a subset of the total database? I'm thinking in something like this:
DB= db.getNodalResult("Displacement", match = MpcOdb.Contains)
DB_subset=DB.getnodes(if id.node == match)
and then iterate for all time steps...
I think in the database methods there are some boolean options (you did something like this in the critical damage virtual result from the same webinar), but I could not figure how to do it myself for my case.
As for the problem with the scripts not showing, I think it is something with OneDrive. If I store the scripts in a not-cloud folder this problem doesn't appear. Also when I record an animation I have to store it out of the cloud. I'm not sure why this happens, but it's ok.