In this paper a contribution to the improvement of accuracy/speed in induction motor optimization is presented. It is based on the reduction of the number of variables involved by first finding similarities between different geometrical shapes in the magnetic circuit. Therefore, a prior filtering of the “shapes” to be computed is performed, allowing the use of accurate optimization algorithms with a reduced number of specimens to be tested. This allows the use of more accurate optimization procedures that conventionally have been limited in their scope due to the time required to simulate every single motor (for instance, finite element analysis).