Hello
I intend to optimize a structure in which some design variables are correlated.
Is it possible to consider the correlation between design variables in the RBDO with UQLab?
Best Regards,
Soheil
Hello
I intend to optimize a structure in which some design variables are correlated.
Is it possible to consider the correlation between design variables in the RBDO with UQLab?
Best Regards,
Soheil
Hi @Soheil_Radfar,
Welcome to UQWorld!
I think @moustapha would know more about this. Let’s see if he’d be around
Thanks @damarginal,
Yes. It would be great if Mr. @moustapha would guide me in this regard
Hi @Soheil_Radfar,
Unfortunately this is not currently implemented in UQLab. If you are using a simulation-based algorithm (e.g. two-level approach with MC in the inner loop), you can easily adapt the codes to define the dependency structure. The input model at each iteration of the optimization is defined in the file “uq_runReliability.m” - Lines 19-40. You can add the lines of codes regarding the dependency structures there for the specific problem you would like to solve.
Please let me know if this is helpful. I will have a look and see if it makes sense to generally add this feature in the module (at least for the simulation-based algorithms, I am not sure when the state-of-the-art approximation methods are used…).
Cheers,
Moustapha
Dear @moustapha,
Thank you for your help.
I need to work on how to implement this in the UQLab. In the coming days, I will present the results of my work here.
Cheers,
Soheil
Dear @Soheil_Radfar
Thanks for your question. I also want to consider the correlation in both of the design and environmental variables by the copula formalism which mentioned here. Unfortunately, I couldn’t implement the @moustapha’s suggestion on changing the:
“uq_runReliability.m” - Lines 19-40
I was wondering if you could share your achievements
It’s worth mentioning that It’s an honor for me to have the developers’ guidance too.
Best regards
Ali
Dear @ali,
I used a different approach (“Bayesian Regression Analysis”) for my study and therefore, I did not follow this solution yet.
Best,
Soheil