Despite having excellent optimization results the real system fails totally – what went wrong?
You have accomplished excellent and promising result when optimizing a complex systems with the help of simulations. Nevertheless the implemented real system fails completely. What went wrong and how can this happen?
The application of numerical optimization methods sometimes animate engineers to exaggerate when running after defined optimization targets. Though optimality and robustness are contrary directions. That way optimality and robustness are the Yin and Yang of a today's development engineers.
Optimality usually comes along with a loss in robustness. Therefore a “good” system design needs balanced portion of both. While primary optimization targets are easy to define most of the time (being quicker, lighter, safer, cheaper,...), robustness can only hardly be comprized by concrete criteria and measures.
When applying arbitrary optimization methods to complex and complicated systems, a dedicated robustness management is necessary to avoid bad surprises in the final implementation of the optimized system. Beside the explicit formulation of proper robustness criteria for a given system, these robustness criteria must be checked permanently or incorporated directly in the system performance rating for the optimization.
Last update on 2016-12-04 by Andreas Kuhn.