Optimal Design of Coreless Axial Flux PM Machines Using a Hybrid Machine Learning and Differential Evolution Method

Coreless stator axial flux permanent magnet (AFPM) machines require computationally intensive three-dimensional finite element analysis (FEA) for accurate performance evaluation, making optimization time-consuming and impractical for large-scale design studies. This paper presents a hybrid optimization approach that integrates differential evolution (DE) with artificial neural networks (ANNs) to accelerate the optimization of coreless AFPM machines. In this method, DEdriven FEA simulations generate a dataset used to train an ANN surrogate model, significantly reducing reliance on direct FEA computations. The effectiveness of this approach is demonstrated through a multi-objective DE optimization, where the ANN's predictions are validated against FEA results. The proposed hybrid method substantially reduces computational cost while maintaining accuracy, providing an efficient solution for electric motor design optimization.