New types of scenario?

MDOScenario is a class for solving an MDO problem from a set of disciplines, variables of interest (objectives, constraints and observables), a design space, an MDO formulation and a driver (either an optimizer or a DOE). Its execution computes (and should return) an optimization result. DOEScenario is an MDOScenario using drivers of type DOE and exporting all the samples from the database, and not only the entries of the database, in order to include duplicated samples ignored by the database.

I’m finalizing an MR about a new class to evaluate a collection of discipline using a DOE, without the optimization notions (i.e. no objective, no constraint, …)
Its execution will return the samples. There will be an option to execute the disciplines either sequentially or using an MDA. This class will be the equivalent of the discipline-free EvaluationProblem.

From these three classes, we can propose a definition:

A scenario is created from a collection of disciplines and a design space (i.e. input space of interest) and executed from an algorithm; its execution returns a result.

Would it be interesting to go further and propose new types of scenarios matching this definition? Either very new features or existing features for which a scenario API would be more user-friendly?

Hi,

Just a thought: in a DOE, wouldn’t it still be relevant to retain the notion of constraints? I’d think being able to filter the results by validity would still be useful, as would being able to see which constraints are violated and by how much.

Hi,

Yes, you’re right. In this case, you can use an MDOScenarioas an MDOScenariomanages constraints and supports DOE algorithms.

In the next release, MDOScenariowill still be there for solving an MDO problem or sampling its quantities of interest, namely objectives, constraints and observables. DOEScenario will probably be removed, asMDOScenariodoes the same job. In addition, there will be a class EvaluationScenarioto sample observables, without having to worry about optimization concepts.