Improved neural network model for induction motor design

Abstract

An improved model of the artificial neural network for analysis and design of induction motors is presented. Parameters of the machine equivalent circuit are calculated using finite element method for a given motor geometry. The training of the neural network model is based on a decoupled system between geometrical variables and circuit parameters. This method efficiently improved the training and performance of the neural network model which can be used to predict machine performance and solve design optimization problems. Index terms-induction machine module, neural network, finite element method. H. C. J. de Jong, AC motor design with conventional and converter Haykin, Neural Networks, A Comprehensive Foundation. Macmillan

Characteristic analysis of 5-phase hybrid stepping motor considering the saturation effect

“…Due to the fact that FEM needs much computing time, the multi-layer feed-forward neural network is employed to save computing time for the circuit parameter [5].…”

“…However, it is not always accurate enough but flexible and timesaving and its accuracy can be improved by using the precise lumped parameters with the nonlinear characteristics. As one of the numerical methods, Finite Element Method (FEM) is a powerful method that provides more accurate analysis results [4], [5], accounting for the nonlinear characteristics and the complex geometry. But it has disadvantages of excessive demand for computational time and resources.…”

Electric machines as metamaterials: Induction machine design using planar layer models

“…Recently, FEM and various other numerical methods have been used to model induction machine performance for design purposes [5]. However, owing to the computational intensity of these procedures, their effectiveness as standalone design tools is limited [6].…”