Abstract: Axial flux permanent magnet (AFPM) motors, with their flat structure and high torque density, have attracted significant attention in cutting-edge fields such as electric vehicles and drones. However, to further break through their performance ceiling, rotor design is a critical variable. This article starts with the flux-focusing principle of the Halbach array and then explains the improved design of the dual-skewed pole structure. It moves into the frontier of computer-aided design, examining how multi-objective genetic algorithms and metaheuristic methods achieve Pareto optimality in motor design. Finally, it focuses on the near-net-shape forming process of soft magnetic composite (SMC) materials and discusses how this technology helps bridge the “last mile” from engineering prototypes to mass production of axial flux motors.
I. Halbach Array and Dual-Skewed Poles: “Fusion” and “Shaping” of the Magnetic Field
The performance ceiling of an axial flux motor largely depends on the quality of the magnetic field distribution produced by the permanent magnets on the rotor side. The traditional surface-mounted permanent magnet (SPM) structure is simple, but its inherent drawback of divergent magnetic flux lines leads to limited air-gap flux density and high leakage flux.
The Halbach array offers a nearly ideal solution. It is a special arrangement of permanent magnets – the magnetization direction of adjacent magnets is sequentially rotated by 90°, so that the magnetic field is enhanced on one side of the array and almost completely cancelled on the other side, achieving a self-shielding effect. In more intuitive terms: in a conventional magnetic circuit the flux lines diverge symmetrically, while the Halbach array “confines” the flux lines to the working air-gap side, realizing efficient flux focusing. Experiments have shown that in axial flux motors employing a Halbach array, torque density can be increased by up to 28% and cogging torque reduced by 65%.
However, the Halbach array also faces challenges in practical rotor design: although the sinusoidal quality of the air-gap flux density is improved, torque ripple – especially cogging torque – remains a major bottleneck for smooth operation. The introduction of dual- skewed pole magnet technology is a precise intervention targeting this pain point.
A 2024 research team from Khon Kaen University in Thailand, publishing in IEEE Access, proposed an innovative TORUS axial flux motor with a skewed Halbach array. By arranging the permanent magnets in a skewed configuration (forming dual-skewed poles), the improved motor, compared to a baseline, showed a 4% increase in back-EMF and a 9.3% reduction in cogging torque under no-load conditions; under load, average torque increased by 8% and torque ripple decreased by 7.8%. These improvements can be attributed to the synergistic enhancement of flux-focusing and flux cancelling effects – the skewed structure extends the degree of freedom for magnetic field regulation in space, effectively suppressing harmonic components of the air-gap flux density.
Other studies have confirmed that for axial flux motors with soft magnetic composite cores, further torque enhancement can be achieved by analytically optimizing the axial magnetisation coefficient (optimal value ~0.82) of a two-segment unequal-width Halbach array. More recent results go even further: a 2025 study published in Scientific Reports adopted a dual-skewed Halbach array double-sided axial flux permanent magnet motor and, through multi-objective genetic algorithm optimisation, achieved a 7.8% increase in average torque and a significant reduction in torque ripple.
II. The “Ace Weapon” of Computer-Aided Design: Multi-Objective Genetic Algorithms and Metaheuristic Methods
If the Halbach array answers the “what to do” question, then modern optimisation algorithms answer the “how to do it optimally” question. For axial flux motors, design variables such as rotor geometry, magnet dimensions, magnetisation angle, and skew angle are coupled in complex nonlinear ways, and traditional single-parameter sweep or trial-and-error methods have long reached their limits.
Multi objective genetic algorithms (MOGA) are currently the most mature class of solutions. They mimic the “survival of the fittest” and “genetic variation” mechanisms of nature, automatically searching the vast design space for Pareto-optimal solution sets through selection, crossover, and mutation operations. Each point on the Pareto front represents a non-dominated trade-off – none of the objectives can be further improved without sacrificing another.
Specifically, NSGA-II (Non-dominated Sorting Genetic Algorithm with elitism) is the most widely used variant. In a domestic study on a V-shaped interior permanent magnet vernier motor, the combination of a BP neural network surrogate model and NSGA-II achieved more than 10% improvement in both torque and core loss optimisation. At the international frontier, a 2025 study by Liu Huijun’s team in Progress In Electromagnetics Research C systematically demonstrated a multi-objective genetic optimisation process with the dual objectives of maximising output torque and minimising torque ripple. In addition, the combination of genetic algorithms and the TOPSIS method has also been proposed for rotor slot structure optimisation in flat-wire permanent magnet synchronous motors.
Multi-objective genetic algorithms do not work alone. The metaheuristic family plays different roles according to the problem characteristics:
· Particle swarm optimisation (PSO), inspired by bird flocking, excels at global optimisation of continuous variables. In the optimisation of a coreless stator axial-field permanent magnet motor, both GA and PSO have been used to maximise output power per unit permanent magnet volume. Weighted inertia-adjusted PSO has also been applied to structural parameter optimisation of an axial-divided-phase magnetic-levitation switched reluctance flywheel motor.
· Artificial neural networks (ANN) act as surrogate models. Because each finite element simulation (especially 3D FEM) can take from minutes to hours, directly embedding them into the optimisation loop imposes a huge computational burden. Therefore, researchers often train ANN surrogates on high-fidelity FEM data, replacing hour-long simulations with second-level predictions and dramatically improving computational efficiency. In the optimisation of a permanent magnet-assisted switched reluctance motor, a genetic algorithm optimised support vector machine (GASVM) was used together with NSGA-II to achieve multi-objective optimisation.
· Ant colony optimisation (ACO) has also been applied to efficiency optimisation of axial flux motors. In the optimisation of a double-stator single-rotor axial-flux brushless DC motor, GA improved the efficiency from 91.01% to 91.57%, while ACO further increased it to 91.80%.
The combined application of these metaheuristic methods has enabled an overall efficiency improvement of up to about 15% for axial flux motors under real operating conditions – a significant achievement in the face of increasingly stringent industry standards for high-efficiency drive systems.
III. SMC Materials and Near-Net-Shape Forming: “Geometric Freedom” in Rotor Manufacturing
If the Halbach array and multi-objective optimisation solve the “electromagnetic design” challenges of axial flux motors, then soft magnetic composite (SMC) materials together with near-net-shape forming technology are rewriting the rules of “manufacturability”.
Soft magnetic composite is a magnetic material formed by pressing iron-based powder with an electrical insulating binder through a powder metallurgy process. The powder metallurgy process creates an insulating layer between the magnetic particles, effectively reducing eddy current losses; at the same time, SMC exhibits isotropic magnetic properties – a fundamental difference from the anisotropic behaviour of traditional silicon steel laminations. Silicon steel can carry high flux density (saturation ≥ 2.0 T) only in its two dimensional rolling direction, but performs poorly in complex three-dimensional magnetic circuits. SMC, on the other hand, supports true three-dimensional flux path design, making it an ideal material carrier for novel topologies such as axial flux motors that inherently rely on a 3D magnetic field distribution.
More importantly, SMC provides rotor design with an unprecedented degree of manufacturing freedom.
Traditional silicon steel cores must be manufactured through a long chain of processes – stamping, stacking, welding, etc. – with low material utilisation and severe geometric constraints. SMC, using powder metallurgy, allows a single-step moulding of highly complex geometrical features. This is the core meaning of “near-net-shape forming”: a design close to the final shape can be directly realised by pressing in a mould, greatly reducing subsequent machining.
This advantage is particularly evident in axial flux motors. In a 2025 study by the Japan Powder Metallurgy Society, SMC was used to integrally form the teeth and double flanges of a stator, significantly increasing the opposing area between stator and rotor while simultaneously improving electromagnetic performance and manufacturing efficiency. A domestic industry report from October 2025 similarly pointed out that SMC, thanks to its isotropic magnetic properties, low eddy current losses, and support for 3D flux design, is driving axial flux motors toward high performance, low energy consumption, and stable mass production. At current process levels, the consistency of SMC stators has been improved by more than 15%, and the overall yield rate exceeds 96%.
In more advanced applications, SMC is also combined with silicon steel to form hybrid stator structures: the silicon steel carries high flux density (≥ 2.0 T) for 2D magnetic paths, while the SMC handles complex 3D flux. Both materials exploit their respective advantages while reducing eddy current losses and design complexity.
Of course, SMC is not without shortcomings. Its magnetic permeability is lower than that of silicon steel, limiting peak flux density in very low-frequency applications; moreover, its brittle nature makes mechanical strength considerations more important for rotor-side use. Nevertheless, for the complex geometries of stator cores in axial flux motors, the advantages of SMC far outweigh its disadvantages – which is why it is regarded as a key catalyst for accelerating the commercialisation of axial flux motors.
IV. Conclusion: Three Keys, One Mission
From the innovation in magnetic circuit principles (Halbach array and dual-skewed poles), to the restructuring of design methodology (multi-objective genetic algorithms and metaheuristic methods), and finally to the paradigm shift in materials and manufacturing (SMC near-net-shape forming), the design of high-performance axial flux motor rotors is undergoing a profound transformation – from “experience-driven” to “computation-driven + materials-driven”.
The Halbach array focuses magnetic flux to unprecedented levels; the dual-skewed pole structure achieves precise ripple suppression; multi-objective genetic algorithms and metaheuristic methods efficiently locate the Pareto-optimal trade-offs between electromagnetic, thermal, and manufacturing costs in a vast search space; and SMC breaks the three-dimensional constraints of traditional manufacturing, giving mass-production feasibility to complex geometries that previously existed only in academic papers. These three keys come together toward a single goal – without sacrificing performance, to bring axial flux motors into our cars, aircraft, robots, and home appliances at lower cost, with shorter lead times, and with higher reliability.
For engineers and researchers, this is not only a continuous expansion of technical boundaries, but also a window of design-paradigm shift worth seizing.
