Title: Design and optimisation of high-speed permanent magnet motor based on artificial neural network

Authors: Jinxing Zhao; Zekai Dong; Jinfeng Bu; Su Lin

Addresses: School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China ' School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China ' School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China ' School of Mechanical Engineering, University of Shanghai for Science and Technology, Shanghai, 200093, China

Abstract: High-speed permanent magnet synchronous motors (HSPMSM) have problems with skin and proximity effects and more high-frequency loss. Furthermore, if the design optimisation of the HSPMSM is directly based on a finite element model, the calculation time is too long. This paper studies the structural design, modelling, and optimisation method of a 100 krpm-10 kW HSPMSM. A finite-element simulation model considering the skin and proximity effect is firstly established. An optimisation method combining artificial neural network (ANN) and multi-objective genetic algorithm (MOGA) is proposed to perform the optimisation of the motor structure parameters. The ANN models are trained and tested with two datasets calculated with the finite-element model. After optimisation, the AC loss of the windings is greatly reduced from 296 W to 186.4652 W, while the optimisation time based on the ANN replacement model is reduced by nearly 70% compared to the optimisation based directly on the finite element model.

Keywords: mechatronics; high-speed permanent magnet motor; high-frequency winding loss; ANNs; artificial neural networks; MOGA; multi-objective genetic algorithm; mechatronics design and optimisation.

DOI: 10.1504/IJMMS.2024.138136

International Journal of Mechatronics and Manufacturing Systems, 2024 Vol.17 No.1, pp.43 - 68

Received: 03 Aug 2023
Accepted: 02 Mar 2024
Published online: 29 Apr 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article