The use of race simulation models as an alternative to robust design optimisation
The idea of the Race Simulation model comes from the America’s Cup: The most challenging design is the one which wins most of the matches under stochastic perturbative conditions.The design which wins (maybe with a great gap) only one match under the most probable condition usually do not win the tournament. This idea is directly applied to design under perturbative conditions (they are not part of the design itself: AoA, speed,etc.) but can be extended to uncertainties of the input parameters. The main focus of this presentation are the advantages and disadvantages of Race Simulation methodology.
The advantages include no arbitrary weights definition, no increment of the objectives number and no preliminary preference for performance or stability, the result of the tournament is an implicit weighted function of mean and variance of performance. On the other hand the disadvantages consist of the fact that a tournament can be very time consuming (but not as MORDO) because of the reduction techniques of matches per tournament, the use of response surfaces and the fact that the tournament score can not be directly used as the overall optimization objective.