ATC-138 Recovery Assessment Error

[anandtp766]

Hi,
I am trying to carry out an ATC-138 recovery assessment using demands from my input file in the PBE application. I can successfully run the damage and loss assessment part alone without any errors. However, when I attempt to run the recovery assessment as well, I encounter the error shown in the attached screenshot. Could you help me resolve this?
Note: I am running the recovery assessment with default inputs. I have also attached the input .json file for reference.

9SBRBF_4_Submitted.json

[fmckenna]

can you zip up and send me (fmckenna@berkeley.edu) the ~/Documents/R2D/LocalWorkDir/tmp.SimCenter folder.

thanks and sorry
frank

[anandtp766]

I've sent the files. Please let me know if you are able to resolve the issue.

Anand

[zsarnoczay]

Dear Anand,

The error you experienced is almost surely due to a bug in the atc138 Python package. The bug causes an error when you use any of the B1031.011a, b, or c components or the B1031.021a, b, or c components. If you remove those components from your performance model (CMP_QNT.csv file), the analysis should run fine.

This bug has already been fixed in April (see this Pull Request if you are interested: OpenPBEE/Functional-Recovery-Python#26). The team behind the atc138 package has been working on a few more updates and plans to make a new release soon. Once there is a new atc138 release, new SimCenter backend releases will start using that, and this bug will be resolved in PBE.

Until then, you can manually update your atc138 package to the most recent commit by running the command below in your command line:

"C:\SimCenter\PBE\applications\python\python.exe" -m pip install git+https://github.com/OpenPBEE/Functional-Recovery-Python.git@main

Note that this requires you to have Git installed on your system. If you do not have it yet, you'll get an error like this:

Cannot find command 'git' - do you have git installed and in your PATH?

You can install git by running the command below in your command line:

winget install --id Git.Git -e

Then, go back and run the pip install command above, and it should work properly.

Once you update the atc138 package, you should be able to continue working with your existing model without removing any B1031 components.

I hope this helps. Please let us know if you are able to resolve the issue.

[anandtp766]

Dear Adam,

Thank you for your detailed response. The model runs successfully in PBE after removing the B1031.011c component. I also updated the atc138 package using the procedure you described; however, PBE is still unable to run my model with the B1031.011c component included (runs into same error again). I would like to clarify a couple of things:

1. Does the procedure mentioned above update only the atc138 Python package, or does it also enable its usage within the PBE application?
2. Is there a Jupyter notebook or Python script available that can run both the damage and loss assessment and the functional recovery assessment in a connected manner, using inputs similar to those used in the PBE application? Essentially, I am looking for a way to run Pelicun and atc138 in a coupled workflow.

Anand

[zsarnoczay]

Dear Anand,

It's great to hear that removing the B1031 component helped get the simulation running.

I am not sure why updating the package did not help. PBE uses the atc138 package installed under the Python environment that is bundled with the Windows release of the PBE tool. Make sure the Python path under File/Preferences in PBE points to this Python environment: C:\SimCenter\PBE\applications\python\python.exe. Anything you change in that Python environment directly affects the simulations run under PBE. Hence, when you updated atc138, the bugfixes should have been made available to PBE.

I'll prepare a Jupyter notebook for you that runs Pelicun and atc138 using the same input files you already have for PBE. I can share this with you later today.

Adam

[anandtp766]

Dear Adam,

Thank you for the clarification.

The python path is correctly set to the environment inside PBE itself. From what I understand, the following command is not properly updating the atc138 package inside the PBE python environment. Upon inspection, I found that the red_tag.py is not getting updated to the latest fix in GitHub. I will look further into this from my end.

"C:\SimCenter\PBE\applications\python\python.exe" -m pip install git+https://github.com/OpenPBEE/Functional-Recovery-Python.git@main

I appreciate you putting together a Jupyter notebook for running both Pelicun and atc138 together. I look forward to it.

Thank you again for your continued support.

Anand

[zsarnoczay]

Dear Anand,

I'm sorry, that's my mistake, I forgot to put the upgrade flag in the command. Pip saw that atc138 is already installed and did not upgrade it. Please try with this command instead:

"C:\SimCenter\PBE\applications\python\python.exe" -m pip install --upgrade --force-reinstall --no-cache-dir git+https://github.com/OpenPBEE/Functional-Recovery-Python.git@main

Note that I also added force-reinstall and no-cache-dir to make sure the existing atc138 will be overwritten by the one from GitHub.

Please let me know if using this command helps.

Adam

[zsarnoczay]

Dear Anand,

I attached a Jupyter notebook that includes a short guide and allows you to run your analysis using only pelicun and atc138:
PBE_to_pelicun_atc138.ipynb

I'd appreciate your feedback on this notebook. If it works well for you, I'll add it to Pelicun so it's available to all users.

Adam

[anandtp766]

Dear Adam,

I was able to resolve the issue by adding --no-deps flag to the command you provided:

"C:\SimCenter\PBE\applications\python\python.exe" -m pip install --upgrade --force-reinstall --no-cache-dir --no-deps git+https://github.com/OpenPBEE/Functional-Recovery-Python.git@main

When I tried using the above command without the --no-deps flag, it reinstalled atc138 along with all its dependencies. This caused dependency conflicts, as some packages were upgraded to newer versions incompatible with the rest of the workflow (e.g., pandas upgraded from 2.3.3 to 3.0.3 and scipy from 1.15.3 to 1.18.0). These version mismatches in turn introduced new errors in the PBE application during recovery assessment.

Using the --no-deps flag restricted the reinstallation to the atc138 package only, leaving all other dependencies untouched. After this, I was able to run the recovery assessment in PBE with the B1031.011c component included successfully.

Thank you for your help in resolving this. I would also like to thank you for providing the Jupyter Notebook for running both pelicun and atc138. I will try to run my recovery assessment using it and provide feedback.

Anand

[zsarnoczay]

Thank you for sharing the info about the --no-deps flag, that's very helpful!

[anandtp766]

Dear Adam,

I was able to run recovery assessment for my model using the Jupyter notebook you provided. I have listed my feedback and some doubts below. It would be great if you could clarify them.

1. The program ran successfully when all inputs, including optional inputs, were provided. I was able to get results comparable to those from the PBE app in the notebook.
2. However, if tenant_unit_list.csv is not provided in the input folder, the program runs into an error in stage 3 during the run_analysis execution (FileNotFoundError). The default tenant_unit_list.csv is not getting generated in the ATC138 input folder.
3. Is there any provision to specify a "seed" value during the damage and loss assessment as well as the recovery assessment, so that reproducible results can be obtained? I noticed that run_analysis in atc138 accepts seed as an optional parameter, but could not find an equivalent for run_pelicun part.
4. I have user-defined fragility functions for some components, while the remaining components use the FEMA P-58 fragility library. This user-defined fragility is specified in input.json under the DL:Asset:ComponentDatabasePath parameter. Is this sufficient for pelicun to incorporate it, or does it need to be provided in a different form? Is the custom_model_dir parameter in run_pelicun related to user-defined fragility, or something else?
5. When using sample_size=5000, I receive warning similar to the following three times during the run_pelicun step. It does not occur when sample_size=500. Should this warning be a cause for concern, and if so, how can it be resolved?

WARNING in pelicun/tools/DL_calculation.py at line 1043
Columns (0, 1, 2, 3, ..., 224, 225, 226) have mixed types. Specify dtype option on import or set low_memory=False.

6. I am getting the following warning during the atc138 run_analysis step (happens in PBE also). What does it mean? Does it cause the elevator component to be skipped during repair scheduling? Also, does the atc138 module skip all components with simultaneous damage states during repair scheduling? I am using D.10.14.011 traction elevator component in my asset model.

Ignoring simultaneous DS. Consider using alternative elevator components without simultaneous damage states.

Thank you for providing the notebook for running both Pelicun and atc138 together. I believe it will be useful for the wider community as well.

Anand