What is the difference between optimum and robust optimum

178 Views Asked by At

While reading some research papers, I come to know that author has explicitly mentioned robust optimum and non-robust optimum terms. Is there any difference? If yes, can you please explain with an example.

1

There are 1 best solutions below

1
On

In my experience, an optimum is robust if it is reasonably distant from constraint boundaries. Yes, optimization tends to drive the design towards a peak, but if there are constraints in the way it will instead tend to drive the design towards 'corners' where different constraints cross. If you are using optimization to design a physical part of a larger assembly, being up against the constraints is bad: The actual manufacturing process can introduce deviations from the calculated optimum, and if those deviations cross a constraint boundary the part may be unusable. If the optimum is right up against two or more constraints, a lot of parts could have to be thrown away because of that. Robust optimization, on the other hand, finds a design which, while not theoretically optimum, is close to optimum while being a certain minimum distance from all constraints. Manufacturing a part according to a robust optimum automatically build in a lot of tolerance, so the manufacturing defect rate can be greatly reduced. If you have heard the phrase 'Six Sigma', then that's what this is about - reducing defect rates to below one-in-a-million, i.e six 'sigmas' (standard deviations), by careful up-front design.

Of course, the author of that paper might mean something else entirely....