Conference Agenda

Topology optimization of permanent magnet motors is one of the promising methods to find high performance designs. However, it is difficult to simultaneously optimize magnets and iron core geometries, because topology optimization often generates too complicated magnet shapes for manufacturing. To address this issue, this paper presents a new topology optimization method for permanent magnet motors that integrates image processing algorithm, which makes it possible to generate meaningful magnet shapes from the view of manufacturing and determines magnetization derection automatically.

In this paper least square boosting based meta-model optimization Algorithm (LSBMOA) is proposed for the optimal design of surface-mounted permanent magnet synchronous motor (SPMSM) for robot joint. The LSBMOA uses machine learning to shorten the time to find multi-objective optimal (MOO) solutions and improve the performance of meta-model. The conventional MOO algorithms, non-dominated sorting genetic algorithm-II (NSGA-II) and multi objective particle swarm optimization (MOPSO), are stochastic algorithms that iterate on the algorithm by increasing the probability that an excellent solution will be chosen. Therefore, they take a lot of function calls and time to find the pareto front set. The proposed algorithm reduces the pareto front set derivation time by directly adding a solution to the objective function area to overcome the limitations of conventional algorithms through a meta-model. Furthermore, the grid subdivision and dynamic sampling process are added to improve the predictive performance of the algorithm. The LSBMOA is compared with NSGA-II and MOPSO to evaluate the performance of the proposed algorithm using two test functions and to confirm its superiority. In order to apply the verified algorithm to the optimal design of a robot joint motor, design variables that affect torque, back-electromotive total harmonic distortion, and cogging torque, which are objective functions to be improved in the initially designed motor, are selected. From the extracted pareto front set, an optimal design is derived by selecting the most suitable solution based on the relative importance of the objective functions.

This article derives the rotor shape of interior permanent magnet synchronous motor (IPMSM) to reduce performance deviation under temperature variation. The overall characteristics of IPMSM in response to temperature variation, such as constant torque curve and maximum torque per ampere trajectories, are analyzed. Moreover, the effects of temperature on efficiency of IPMSM are investigated. As a result, for reducing average torque and efficiency deviation caused by temperature changes, the proportion of reluctance torque needs to be increased. In this article, the improvement design considering temperature variation are conducted by positioning the magnets more vertically compared to the original model. Finally, to validate the theoretical analysis and numerical simulations, a prototype is fabricated and tested.

This paper proposes a novel grain-oriented consequent pole surface-mounted permanent magnet motor (GCSPM) to reduce rare-earth magnet usage. Conventional consequent pole (CP) motors typically replace one pole of the magnet with non-grain-oriented steel (NO), allowing for rare-earth magnet reduction but resulting in lower electromagnetic performance compared to conventional motors. To address this limitation, the proposed GCSPM utilizes grain-oriented steel (GO) to improve electromagnetic performance compared to conventional CP motors. The GO exhibits superior B-H characteristics along the rolling direction. However, GO shows lower B-H characteristics in the direction perpendicular to the rolling direction. Therefore, the proposed GCSPM replaces the south pole magnet with radially aligned GO in regions where the magnetic flux is relatively consistent to achieve higher electromagnetic performance relative to the reduced rare-earth magnet usage. Additionally, skew and magnet chamfering techniques are applied to reduce torque ripple. The average torque and torque ripple characteristics are analyzed using finite element analysis (FEM) and investigated by varying skew, magnet chamfering, and design parameters to improve performance.

A rotor shape with tapering is a widely adopted design technique that sinusoidalizes the air-gap flux to reduce back-electromotive force (B-EMF) total harmonic distortion (THD) and cogging torque. However, as the air-gap length distribution changes, the parasitic capacitance caused by the geometry of the motor also varies, which significantly affects the shaft-to-frame voltage, leading to motor lifespan degradation due to bearing currents. Therefore, this paper analyzes the changes in parasitic capacitance and shaft voltage caused by the application of rotor tapering and proposes a model for their analytical model. The analytical model predicts the parasitic capacitance of the tapered rotor and is validated through electrostatic finitie element analysis (FEA). The validation results show that applying a tapered rotor increases the shaft-to-frame voltage, confirming it as a factor to be considered in the design process. Therefore, the B-EMF THD, average torque, and shaft-to-frame voltage are set as objective functions, and an optimal design is performed using a machine learning-based genetic algorithm (GA). Finally, prototypes of the initial model and optimized model are manufactured, and the results are validated through no-load, load conditions, and shaft-voltage experiments.

In this study, an optimal design method for axial flux permanent magnet motors (AFPMMs) is proposed by intergrating a combined analysis approach that utilizes static three-dimensional (3D) finite element analysis (FEA) and quasi-3D FEA, with a kriging-based optimization algorithm. This approach aims to alleviate the computational burden typically encountered in the design optimization of AFPMMs. The proposed method addresses challenges related to complex magnetic interactions and high computational costs by alleviating FEA time through iterative improvements in 3D model analysis. Simultaneously, the kriging surrogate model efficiently approximates the design space, ensuring precision while significantly reducing computation time. The proposed optimization strategy enhances signicicant performance indicators such as torque density and torque ripple, demonstrating its effectiveness in optimal design of AFPMMs. Moreover, the methodology is broadly applicable to other complex electromagnetic systems, particularly radial flux permanent magnet motors (RFPMMs) with overhang structures that require 3D FEA. This approach offers motor designers and engineers an efficient optimization framework, enabling rapid and accurate derivation of optimal design solutions while minimizing computational time.

This article proposes a novel power trunk locking system using electromagnetic force to remove protrusion of conventional trunk locking system. The proposed system consists of contact plate, magnetic module, gear train, and cinching motor. The electromagnetic force from the permanent magnet in the magnetic module is applied to the contact plate attached to the trunk, thereby maintaining the holding force. In this article, firstly, the principle of a proposed magnetic trunk locking system is introduced. Subsequently, it is defined that the performance of the magnetic module should be evaluated in terms of the locking on force, locking off force, and cylindrical magnet rotational torque. The performance of the prototype magnetic module is analyzed using three-dimensional finite element analysis, and further verified by experimental tests. Finally, design variables for weight reduction of the magnetic module are selected, and weight reduction design is carried.

In this paper, we present a 3-D multi-material topology optimization (MMTO) method to improve the torque density of a surface permanent magnet (SPM) motor. In the proposed MMTO method, a 3-D optimization model is divided into several layers along the axis of rotation. Subsequently, the material distributions and magnetization directions of permanent magnets for each layer are simultaneously optimized. The 3-D optimized SPM motor demonstrates better torque density when compared with the reference and 2-D optimized counterparts since we directly consider the 3-D magnetic properties including the end effect.

This study focuses on topology optimization of a coupled inductor for a multi-phase buck converter, addressing the challenges of achieving high efficiency and compact design. Using NGnet-based topology optimization, the cross-sectional area of the coupled inductor was optimized to achieve high system efficiency and minimize the volume. The optimization results highlight the importance of considering both coil and circuit losses in the design stage, providing a foundation for further optimization in multi-phase systems.

A novel circuit generation method considering circuit topology is proposed based on heuristic optimization algorithm. This method determines the type and value of elements by searching for a continuous space, and it automatically constructs the circuit topology. Given that this method handles real values, it can also be applied to circuits with non-existent elements, such as equivalent circuits, and it can be used as a design aid for electrical equipment. The proposed method is applied to an equivalent circuit that expresses the complex permeability of electromagnetic steel sheets.

This paper investigates robust design optimization (RDO) for surface-mounted permanent magnet synchronous motors (SPMSMs) to mitigate additional harmonic components (AHC) of cogging torque caused by manufacturing uncertainties. A prototype motor is analyzed using finite element analysis (FEA) and experimental validation, revealing discrepancies due to manufacturing uncertainties. Sensitivity analysis identifies significant variables influencing AHCs, which are then used to develop a kriging surrogate model. This surrogate model facilitates computationally efficient optimization to minimize AHCs while maintaining performance. The optimized design is validated experimentally, demonstrating reduced sensitivity to manufacturing uncertainties and improved motor performance.

Electromagnetic applications often involve uncertainties, including material property variations, geometric inaccuracies, and environmental fluctuations. Such factors can impact design performance, causing deviations from target specifications. Robust optimization enables to obtain design point almost unsensitive to that uncertainties. However, this approach often requires extensive repetitive computations, making the process time-consuming and computationally intensive. To address this challenge, this work proposes the use of neural network surrogate models for robust multi-objective optimization, which can significantly reduce computational time.

This study introduces a novel approach for designing manufacturable 3D coils through topology optimization, utilizing a computationally efficient stream function method. Traditional stream function methods, encumbered by computationally intensive matrix operations like singular value decomposition and limited to designing current density distribution over two-dimensional curved surfaces, are significantly improved upon. We extend the applicability of the stream function method to three-dimensional space, enabling the design of practical, three-dimensional coil configurations. The optimization of the curved target coil surfaces leverages the Covariance Matrix Adaptation Evolution Strategy coupled with NGnet. Furthermore, we accelerate the stream function computations by incorporating the Adaptive Cross Approximation algorithm. This synergistic combination of techniques significantly reduces the computational cost, facilitating the efficient design of complex, manufacturable three-dimensional coil geometry.

This paper introduces a reduced normalized Gaussian network (NGnet) method for topology optimization, which effectively reduces the optimization complexity while preserving the advantages of NGnet such as coordinate system independence and no gray scale generation. In this method, the basis functions are derived by truncated eigendecomposition of the kernel matrix composed of the NG functions. Comparative optimization results demonstrate that this method effectively balances the complexity and diversity of topological structures. It provides an effective approach for achieving more manufacturing-friendly structures.