Dear NLR team,
Does WEIS include RBDO (Reliability Based Design Optimization)?
Best Regards,
Riad
There is no precise definition of RBDO, so I’ll take a stab at answering based on my current understanding of the term. One can include long-term component fatigue as a constraint or objective in design optimization in WEIS. This is best supported for the blades, shaft(s), and tower. To a lesser degree, it can be done for a floating platform as well. Uncertainty in the environment can be sampled through one of the standard Design Load Cases, or you can customize your own. If you wanted to explore uncertainty quantification for material properties or other parameters, there is no direct support for that within WEIS and you would have to create your own outer loop wrapper for those studies.
1 Like
Dear @Garrett.Barter,
Thank you for your answer. I meant by RBDO designing component under probabilistic constraint. Basically, the RBDO has two loops: the outer loop that runs the optimization and the inner loop that perform the reliability analysis (lets say perform Crude Monte Carlo to see whether the solution given by the outer loop respects the constraint of the probability of failure).
On the other hand, since i am planning to start WEIS soon, I want to check whether it requires a lot of computational ressources? For example, how long it will take to redesigning the tower of a 5 MW barge-type FOWT in a deterministic way (as you always do it) using parallel computing on 8 cores ?
If you could gimme an approximation of the time required.
Thank you in advance,
Best Regards,
Riad
Hello Riad,
Yes, for probabilistic design, you will likely have to create your own outer loop that drives WEIS. We also use WEIS with an outer loop wrapper for various internal studies and have therefore created some examples of what that might look like. The following WISDEM driver does a customized design of experiments, but there would certainly be some overlap with your simple Monte Carlo example (there are built-in OpenMDAO design of experiment drivers, but those would be less helpful):
For your other question around deterministic timeline, I have two comments. First is that WEIS is meant for multifidelity scaffolding, so for your tower optimization atop a barge platform, it is probably best to start with a frequency domain optimization in RAFT with larger bounds for your design variables and then take the finished output there and run it through an optimization with OpenFAST using narrower bounds. This will be a quicker path to a final solution. The deterministic time for a single optimization depends on whether you are running with RAFT or OpenFAST, how many design variables, how many constraints, how many DLCs (with however many wind speeds, wave conditions, turbulent seeds, etc). Most practitioners start with simple examples and build up complexity slowly as they gain confidence in the process and a feel for the computational timelines. This is usually a good approach for any optimization problem.
Cheers,
Garrett
1 Like