Projection-based objective space reduction for many-objective optimization problems: Application to an induction motor design

In this paper an attempt at devising an efficient strategy for solving many-objective optimization problems through objective space reduction has been made. An approach similar to the Targeted Projection Pursuit (TPP) technique is used in order to decompose the optimization problem into one or more lower dimensional problem(s). The design optimization of a double squirrel cage induction motor with five objectives is carried out using this approach. The results show that adopting the proposed strategy can lead to obtaining at least parts of the Pareto optima, while trying to solve the problem with equal computational budget in the original objective space using conventional multi-objective evolutionary algorithms does not lead to proper convergence.