Research in Natural Hazards Engineering > Damage & Loss (PELICUN)

Uncertainty in consequence models

(1/2) > >>

joramh:
Hello Adam,

Thanks a lot for the new updates and versions of Pelicun (specially the tutorial), it has been really helpful. I’ve been working on the tutorial you posted on youtube for a while now and I have a couple of questions regarding the definition of uncertainties.

1.- How can I specify uncertainty in the consequence models? I tried doing this by specifying a family and a Theta 1 value in the collapse/replacement consequences (additional_consequences dataframe in the tutorial), but when I do this, I notice that I still get a deterministic value for all the collapse/replacement realizations. In addition, some of the default FEMA P-58 components have already uncertainty defined and when the analysis is run, all the consequences (replacement costs) are still deterministic. Do you have any recommendations? I’m probably missing something here...

2.- It is mentioned in the tutorial that Pelicun supports many distributions. Does it support skew normal distribution as well? If so, how can I define it? My initial thought was to have Theta_0, Theta_1 and Theta_2 that represents shape, loc and scale of my distribution but I don’t know if this is possible.

Any help will be greatly appreciated! Thanks a lot!
Best,
Andrés Ramos.

adamzs:
Hi Andrés,

Thanks for reaching out; I appreciate your feedback and questions on the new version of Pelicun!

1) This might be a bug, let me double-check and get back to you shortly. Until then, make sure you run the tutorial notebook with version 3.1b4 - there were a few other bugs caught earlier that I've fixed already in that latest beta release.

2) Pelicun does not support skew-normal yet, but it shouldn't be too hard to add it to the script. You'd need to add a few lines of code to the uq.py and model.py modules. Let me know if you're interested and I am happy to provide more information.

Adam

joramh:
Hey Adam,

Thanks a lot for getting back to me. I've updated now to the 3.1b4 version.
Yes, I'm interested in incorporating skew normal distributions so any information you could give me will be really helpful!

Best,
Andrés.

joramh:
Hey Adam,

Apologies if you're super busy, but do you think you could give me more information about this please?
Thanks!
Andrés.

adamzs:
Hi Andres,

I'm sorry for the delay. I'll get back to you with a detailed description tomorrow.

Adam

Navigation

[0] Message Index

[#] Next page