Search

This paper presents the machine learning-based detection of foreign metal object for the wireless power transfer device including differential coils. To test the proposed method, the differential voltages are computed using finite element method for about 1500 cases with and without an aluminum cylinder at driving frequency of 85 kHz considering misalignment between the primal and secondary coils. It has been shown that gradient boosting decision tree and random forests classifier have the accuracy over 90% when input voltages and differential voltages are inputted together.

A novel automatic design method for permanent magnet (PM) motors using a Monte Carlo tree search is presented. The optimal motor structures are determined through a tree search, in which the motors with different numbers of poles, current phase angles, PM configurations, and numbers of PMs are simultaneously considered. At the leaf nodes, parameter and topology optimizations are performed to obtain the optimal material shape and distribution. The proposed method was applied to the optimization of a 24-slot motor. It was shown to be effective in finding the optimal motor structure and geometry to maximize the average torque while considering iron loss. The proposed method can be applied not only to the design of PM motors but also to many types of electric apparatus and other systems.

This article proposes a novel 2.5-D multi-phase topology optimization method using a Gaussian basis function for permanent magnet motors. The design region in the rotor was sliced into cylindrical layers; the 2-D topology optimization was performed for each layer such that the average torque was maximized, while the torque ripple was suppressed to the maximum possible extent. The proposed topology optimization could determine the rotor core and magnet shapes, as well as the magnetization direction. It was shown that the optimized 2.5-D topology optimization led to better torque performance when compared to conventional 2-D optimizations.

This study proposes a method for wireless power transfer systems to identify the existence of foreign metal objects and simultaneously predict the misalignment distance between the primary and secondary coils. The proposed method is based on a neural network (NN) trained using electromagnetic field simulations. The training data for the NN consist of the differential voltages in the detection coils, together with the input voltage of the primary coil. Although the metallic objects and coil misalignment induce confusing voltages, the trained NN exhibits over 90% accuracy for the validation dataset, and mean prediction errors of less than 1 mm for the misalignment distance and ground clearance variance.

This paper proposes a new method for accurately predicting rotating machine properties using a deep neural network (DNN). In this method, the magnetic field distribution over a cross-section of a rotating machine at a fixed mechanical angle is used as the input data for the DNN. The prediction accuracy of the torque properties of an inner permanent magnet (IPM) motor for the CNNs trained by the magnetic flux density distribution and material configuration is compared. It is shown that the proposed method facilitates a more accurate prediction of machine performance than a conventional method in which the cross-sectional image of a rotating machine is input to the DNN. Furthermore, the DNN learned by the proposed method is applied to the topology optimization algorithm. Topology optimization can be effectively accelerated because the number of analyses by the finite element method can be reduced using the proposed method. The total computing cost is reduced by 52.5% compared with conventional optimization without surrogate models.

This article presents a fast population-based multi-objective optimization of electromagnetic devices using an adaptive neural network (NN) surrogate model. The proposed method does not require any training data or construction of a surrogate model before the optimization phase. Instead, the NN surrogate model is built from the initial population in the optimization process, and then it is sequentially updated with high-ranking individuals. All individuals were evaluated using the surrogate model. Based on this evaluation, high-ranking individuals are reevaluated using high-fidelity electromagnetic field computation. The suppression of the execution of expensive field computations effectively reduces the computing costs. It is shown that the proposed method works two to four times faster, maintaining optimization performance than the original method that does not use surrogate models.

Topology optimization (TO) has been formulated as a quadratic unconstrained binary optimization (QUBO) problem that can be solved by Digital Annealer (DA) and quantum computers, which can realize massive parallelization. It is shown that the 3-D TO of a permanent magnet (PM), with 600 unknowns, is successfully performed by the DA with the aid of the finite element method (FEM) for field computations. Moreover, it is shown to perform TO of a magnetic core, with 758 unknowns, in such a way that the QUBO and field problems are iteratively solved to determine the optimal solution in which the magnetic field and optimal core structure are self-consistent.

This article proposes a novel topology optimization approach for permanent magnet synchronous motors (PMSMs) driven by a pulsewidth modulation (PWM) current control system operating in different modes such as maximum torque per ampere (MTPA) and weakening flux controls. The cost function consits of iron and copper losses, which are computed by considering the spatial and time harmonics in the current generated under these conditions. It is demonstrated that the total loss of the optimized model is reduced compared with that of the conventional model in the operating region.

In this study, we propose a fast topology optimization (TO) method based on a deep neural network (DNN) that predicts the current-dependent motor torque characteristics using its cross-sectional image. The trained DNN is shown to provide the current condition that provides the maximum torque under the assumed motor control method. The proposed method helps perform TO with a reduced number of field computations while maintaining a high search capability.

Ferromagnetic structures, particularly the anti-eddy plate, in a bi-planar permanent-magnet-type low-field (0.05 T) magnetic resonance imaging (MRI) brain scanner can distort the gradient field in the target region. This study aims to provide a new gradient coil design method that reduces ferromagnetic influences on gradient field linearity. Thus, a simplified model of electromagnetic (EM) structures of the permanent-magnet-type MRI scanners was established. By using precise analytical proof, the anti-eddy plate was reduced to a homogeneous magnetic plate. The overall effects of the EM structures, which can be represented by bi-planar magnetic plates, were evaluated. In sequence, the image magnetic dipole was first introduced to show the effects of anti-eddy plates were added to the conventional equivalent magnetic dipole (EMD) approach. A novel equivalent image magnetic dipole (EIMD) method was proposed to build the gradient coil pattern. The effect of ferromagnetic materials was predicted throughout the gradient coil design phase using the proposed method, and a high-linear gradient field was generated under real working conditions. The computational and experimental results showed that the gradient coil was linear when ferromagnetic structures were present. The effectiveness of the proposed method was demonstrated by comparing T1-weighted images of the conventional method to those of the proposed method. The proposed method reduced image distortion caused by nearby EM structures in bi-planar permanent-magnet-type low-field MRI systems and provided an effective and concise solution for gradient coil designs.