Rapid co-kriging based multi-fidelity surrogate assisted performance optimization of a transverse flux PMLSM

Abstract

Conventional struggles to achieve optimal design for transverse flux machines with many design parameters is still computationally an expensive task. By employing surrogate assisted indirect optimization methods, the task at hand is accelerated. However, for large design parameters vector, initial finite element analysis samples required to generate those surrogate models is high. Coupled with the fact, that use of three dimensional FEM is indispensable for a transverse flux machine, optimal design demands high computational resources and time. To reduce design time without loss of accuracy, this paper proposes to exploit the correlation between a low fidelity (Equivalent 2D-FEM) and a high fidelity (3D-FEM) model and generate a co-kriging based multi-fidelity surrogate which is then employed for optimization. The method is applied to a hybrid transverse flux PMLSM. Compared to conventional surrogate based methods, considerable reduction in design time is achieved.