Purpose The purpose of this paper is to propose a methodology to seek the optimal topology of electromagnetic devices using the density method while taking into account the non-linear behaviour of ferromagnetic materials. The tools and methods used are detailed and applied to a three-dimensional (3D) electromagnet for analysis and validation. Resulting topologies with and without the non-linear behaviour are investigated. Design/methodology/approach The polynomial mapping is used with the density method for material distribution in the optimisation domain. To consider the non-linear behaviour of the materials, an analytical approximation based on the Marrocco equation is used and combined with the polynomial mapping to solve the problem. Furthermore, to prevent the occurrence of intermediate materials, a weighted sum of objectives is used in the optimisation problem to eliminate these undesired materials. Findings Taking into account the non-linear materials behaviour and 3D model during topology optimisation (TO) is important, as it produces more physically feasible and coherent results. Moreover, the use of a weighted sum of objectives to eliminate intermediate materials increases the number of evaluations to reach the final solution, but it is efficient. Practical implications Considering non-linear materials behaviour yields results closer to reality, and physical feasibility of structures is more obvious in absence of intermediate materials. Originality/value This work tackles an obstacle of TO in electromagnetism which is often overlooked in literature, that is, non-linear behaviour of ferromagnetic materials by proposing a methodology.

Purpose This paper aims to deal with a performance comparison of an 8/6 radial-flux switched reluctance machine (RFSRM) and an axial-flux switched reluctance machine (AFSRM), presenting equivalent active surfaces. Design/methodology/approach An axial machine was designed based on the equivalent active surfaces of a radial one. After estimating the machine inductances with a reluctance network, finite elements numerical models have been implemented for a more precise inductance determination and to estimate the electromagnetic torque for both machines. Finally, the AFSRM was thoroughly examined by analyzing the impact of some geometric parameters on its performance. Findings The comparison of the RFSRM and AFSRM at equivalent active surfaces showed that the obtained axial machine is more compact along with an improvement in the electromagnetic torque. Practical implications The equivalent AFSRM is more compact, therefore more interesting for transport and on-board applications. Originality/value The RFSRM and AFSRM performance comparison using the same active surfaces has not been done. Moreover, the AFSRM presented has a rare design with no rotor yoke and where the rotor teeth are encapsulated in a nonmagnetic structure, allowing a more compact design.