Summary

Knowing the influences of each parameter, DOE allows us to:

• reduce costs by relaxing tolerances on those parameters with little influence,

• know which parameters are really influential, and thus to simplify or establish a hierarchy for the optimization phase,

• obtain a Response Surface Model of a costly objective function by maximizing precision according to the maximal number of experimental runs planned in the calculation budget.

Optimization provides the best possible values (with a level of robustness that often needs to be verified afterwards).

• It can be performed, at least partially, on a surrogate model obtained via RSM (please note that in this case, the validity of the solution must be checked against the original model, especially for the constraints).

RSM via Monte-Carlo analysis or DOE allows us to:

• visualize graphical approximations of objective functions (and therefore to emphasize tendencies or understand parameter interactions),

• get surrogate mathematical models of costly functions to speed up the optimization phase,

• evaluate the robustness of the objective functions and, knowing the governing parameters, to increase (if necessary) the level of robustness by altering the optimal set of values, while keeping the commonly resulting performance loss under control.