This paper aims to propose a data-driven approach to determine the design guidelines for low-frequency electromagnetic devices.
Two different devices, a core-type single-phase transformer and a motor-drive system, are used to show the usefulness and generalizability of the proposed approach. Using a finite element solver, a large database of design possibilities is created by varying design parameters, i.e. the geometrical and control parameters of the systems. Design rules are then extracted by performing a statistical analysis and exploring optimal and sub-optimal designs considering various targets such as efficiency, torque ripple and power factor.
It is demonstrated that the correlation of the design parameters influences the way the data-driven approach must be made. Also, guidelines for defining new design constraints, which can lead to a more efficient optimization routine, are introduced for both case studies.
Using the proposed approach, new design guidelines, which are generally not obtainable by the classical design methods, are introduced. Also, the proposed approach can potentially deal with different parameter–objective correlations, as well as different number of connected systems. This approach is applicable regardless of the device type.
