RBDO with QMC: Two-stage Enrichment Scheme

Dear @moustapha

In this topic I want to mention some parts of your precious work entitled “Surrogate-assisted reliability-based design optimization: a survey and a unified modular framework” and benefit from your comments/clarifications regarding the QMC-based RBDO problem:
In the case-3 analysis, the following statements are mentioned in your paper with Prof. Sudret:

  • “we consider a two-stage enrichment scheme”

  • “The idea behind is to locate the limit-state surface without necessarily focusing on extreme accuracy”

  • “The remaining epistemic uncertainty due to surrogate modelling is taken care of in the second stage where optimization is coupled with enrichment”

  • “case no. 3 does not exactly illustrate the proposed framework and is qualified here as gray-box because of the coupling between optimization and surrogate modelling.

I was wondering if the implemented QMC-based RBDO in UQLab benefits from the mentioned two-stage enrichment scheme or not? In other words, It seems the grey-box modelling is already implemented in the RBDO module of UQLab or this two-stage configuration is not supported in UQLab yet.
Last but not least, as always, I really appreciate any help from RSUQ and other community members.

Best regards

1 Like