Automated Preliminary Design of Induction Machines Aided by Artificial Neural Networks

Abstract

$A$ method for an automated preliminary design of induction machines with squirrel-cage rotors is presented. The speciality of this method is the use of artificial n eural networks. Based on input parameters like voltage, frequency, number of pole pairs and output power a motor design is examined. The motor design can be influenced witho ptional p arameters. The design process uses artificial neural networks to determine several geometric quantities e.g. the stator slot geometry as well as the rotor slot geometry. Moreover, different numbers of rotor slots are considered during the design process to evaluate the parasitic behavior with the corresponding number of stator slots. To show the functionality and clarify the benefits o f t his a pproach, two examples are examined and compared to commercially available machines. The automated preliminary design can be a starting point for further optimization.