High-efficiency analysis and optimal design of motor performance accelerated by surrogate-assisted models

Finite element analysis (FEA) is a commonly used method in motor design, offering relatively high computational accuracy. However, its high computational complexity and time-consuming nonlinear iteration is not conducive to the rapid and efficient design of motors. In this paper, an accelerated analysis and optimization framework for electric motor design based on surrogate-assisted model is presented. By training a surrogate model, the magnetic permeability distribution in each region of the motor can be predicted. These predicted values are then used as initial values in the motor performance analysis model, such as the high-precision reluctance network analysis (RNA) model, for iterative solutions. This significantly reduces the number of model iterations and accelerates the convergence process. Compared with other surrogate model methods that directly predict motor performance, the proposed method avoids the negative effect of prediction errors from the surrogate model on the results, achieving fast and efficient motor performance analysis. In this paper, the performance analysis of a surface-mounted permanent magnet motor (SPM) is taken as an example. The research result show that the high-precision RNA model accelerated by the surrogate model can reduce the solution time by 82.3%.