The motivation for using modeFRONTIER^TM and a general description of the whole design process can be found in the chapters of Why do design optimization while in this chapter a pragmatic description of the tools offered by modeFRONTIER^TM and suggestions on how to use them is given.

Any optimization process can be divided into different phases:

The first phase is performed using Design of Experiments techniques, the second phase using statistical multi-objective algorithms (MOGA-II,MMES, MOGT), the third phase using single-objective optimizers (Simplex, B-BFGS, DES).

Between the second and the third phases a decision process take place where the many objectives are collapsed into one using a transfer function (MCDM or other approaches).

As the path to the optimum is, in many cases, quite complex the user should interactively try different strategies until he has effectively reached the optimum or is satisfied with the solution found. Typical optimization problems are usually solved by means of "hill climbing" procedure possibly based on local gradients of a stated cost function. The typical drawback of this approach is the fact that the search for improvements is done efficiently but is done locally. On the other hand probabilistic optimization techniques can be used to examine a large but discrete configuration space in order to find a good solution possibly close to the global optimum. Typical optimization are more accurate but less robust than probabilistic optimization techniques.

As any algorithm needs an initialization set of points, these are taken from the predefined DOE. The number of initial points differs according to the selected algorithms and may range from one single point for the quasi-Newton method B-BFGS to many points in the case of MOGA-II.

To enter the Design Schedulers panel, double click on the Scheduler node identified by the icon .

The window is divided into three parts:

A specific algorithm can be selected by a left mouse clicking on the name in the list and with the Apply Schedulers button.

Default values for the algorithm are resumed by the button Default Parameters.