Abstract
Step-skew is a common technique for eliminating the cogging torque of a target harmonic order in permanent magnet synchronous motors (PMSMs). However, when step-skew is applied to the rotor, the cogging torque of the target harmonic order is not completely eliminated due to 3-D leakage flux. Therefore, the 3-D leakage flux should be considered in designing a PMSM with step-skew for cogging torque reduction. The most accurate way to consider the 3-D leakage flux is to perform 3-D finite element analysis (FEA), but it has the disadvantage of high computation time. To resolve this challenge, this article proposes a design method that utilizes transfer learning to reduce the time for 3-D FEA while maintaining accuracy. Through the proposed method, a large amount of 2-D FEA-based data and a small amount of 3-D FEA-based data are used instead of a large amount of 3-D FEA-based data, with similar accuracy as using a large amount of 3-D FEA-based data, and the computational time is highly reduced. Finally, a prototype is fabricated and tested to verify the validity of the proposed design method for cogging torque reduction.
Original language | English |
---|---|
Article number | 8204905 |
Journal | IEEE Transactions on Magnetics |
Volume | 59 |
Issue number | 11 |
DOIs | |
State | Published - 1 Nov 2023 |
Keywords
- 3-D leakage flux
- cogging torque
- deep neural network (DNN)
- permanent magnet synchronous motors (PMSMs)
- step-skew
- transfer learning