Non-linear model of synchronous reluctance motor with neural networks

The application of artificial neural networks (ANN) in modeling the synchronous reluctance motor (SRM) is studied. Motor geometry, used materials and excitation level dictate motor performance. Connections between these parameters were investigated due to some simplifications of model multidimensionality. To overcome model complexity proper ANN topology and data presentation to ANN were established Final form of non-linear model was achieved by joining all trained ANN. Model output results define the magnetic flux density distribution along the air-gap for both d- and q-mode excitation and specific desired motor geometry. Comparison between ANN model results and finite element calculation confirms the use of the non-linear model thus established for further investigation of SRM performance.