Uncertainty identification method using kriging surrogate model for industrial electromagnetic device

Abstract

To obtain accurate results from various probabilistic design optimization applied to electromagnetic devices, the uncertainty in the motor should be first identified correctly. This paper presents an efficient uncertainty identification method by using finite element analysis and experimental data. Kriging surrogate model is employed to reduce the computation and maximum likelihood estimation is used to identify the probability distribution of uncertainty and find its parameters. The proposed method is applied to identify the uncertainties in a surface-mounted permanent magnet synchronous motor that cause the cogging torque.