Deep Neural Network를 이용한 매입형 영구자석 동기전동기의 파라미터 예측

Abstract

This paper, the prediction method of parameter for interior permanent magnet synchronous motor using Deepneural network (DNN) is proposed to exactly predict the motor parameter. For the DNN surrogate model to prediction theparameters well and to minimize the computational cost, the experimental points should be evenly distributed within the designarea. However, as the design variable increases, the number of experimental points must be increased. Thus, a highcomputational cost is required. Therefore, in this paper, to reduce the calculation cost, the motor parameters according to theshape change of the motor are predicted through the following two steps. First, the motor parameters were predicted using theDNN surrogate model for the change in the stator outer diameter and split ratio. Then, the motor parameters were calculatedmathematically for the changes in the stack length and the number of series turns per phase.