Optimization of instantaneous torque shape of PM motors using artificial neural networks based on FE results

Abstract

A method that combines artificial neural networks (ANN) and finite-elements method is introduced to estimate the instantaneous torque of two classes of permanent magnet motors. Using parallel supervised multilayer neural networks, the geometrical parameters of the motors are mapped to the developed instantaneous torque. The obtained results show a close agreement between the ANN estimated torques and those computed using the finite element model. The method is then used in an iterative designing process to minimize torque ripple of a permanent magnet motors