Conference Agenda

In the boundary element method (BEM), we deal with the direct and indirect boundary integral equations (BIE(s)). The indirect BEM

has less degree of freedom compared with the direct BEM, and it is commonly used for a long time. It has been numerically proven that

the double-layer approach (DLA) possesses superior features to analyze static electromagnetic problems. However, the indirect BEM

encounters so-called cancellation-error as a drawback without exception. It is well known that the Surface Charge Method (SCM)

encounters the cancellation-error when evaluating the electromagnetic fields in the magnetic materials while the DLA does when evaluating the

fields in the air. In order to avoid the cancellation-error in the DLA, we propose a unified approach available to any static electromagnetic

model by preparing a post-processing with the help of the direct BEM. In addition, we demonstrate how identical formulation for DLA

can be used for both electrostatic and magnetostatic class of problems, independent if the problem is linear or nonlinear.

In a number of applications, a set of complex magnetic sources can be modelled with a simple current distribution able to provide, in

an assigned region, the same field within the required accuracy. The knowledge of the field in a number of points in the region of interest, given for example by a diagnostic system, can be used for the estimation of the equivalent current. The approach, successfully proposed in the past for Tokamak devices to model the plasma current effect, is here extended to treat all the active sources. The numerical treatment of the equivalent current is also discussed in a finite dimension space with a scalar stream function.

To speed up the eddy current computation in case of moving conductors with volume integral formulations of the EFIE, iterative techniques and Fast Multipole Methods (FMM) are applied to take into account the mutual effects between the conductors.

Magnetic microwire (MW) materials, with microwires embedded in a host polymer, can be used to guide the magnetic flux along the wires. Due to the microscale diameter of the individual wires, the eddy current losses are very low, which makes MW materials an attractive alternative for the design of conducting screens and cores of electrical machines. The benefits of such designs are demonstrated in the paper via numerical simulation and homogenization.

FFT-based homogenization models are able to describe the local and macroscopic responses of composite materials from digital microstructure images. But these models, when used for quasistatic applications, are classically based on scalar potential formulations since it does not require a gauge. This abstract presents the dual formulation based on vector potential for magnetostatic problems. Results from both formulations are presented and compared.

A PEEC formulation is applied for the computation of Joule losses and currents in the wires composing the screens of the phase conductors in a submarine tripolar cable. In order to model the infinite extent of the cable and to reduce the computational burden, simmetries of the geometry are exploited. A good agreement is obtained with brute force FEM simulations, with a dramatic reduction of computational times and memory requirement.

This paper combines the convolutional neural network (CNN) and the domain decomposition method (DDM) to evaluate the magnetic field of electrical devices. Taking a model considering the magnetic material anisotropy as an example, the computation results of the finite element method (FEM) are used as the ground truth, the CNN is introduced to train and predict the magnetic field distribution under different geometries. To address the problem of partial information loss due to nonlinearity between geometric parameters and the maximum value of magnetic flux density. DDM is utilized to divide the whole database into 4 subsets and train the CNN on each subset separately. The numerical experiments illustrate that the training performance on each subset is better than training on the entire dataset. It is proven that the introduction of DDM can improve the performance of deep learning.

Electromagnetic force is applied to many electrical machines such as the magnetic levitation systems, actuators, motors and so on. The high accuracy of electromagnetic force calculation is required to accurately analyze the dynamics of electrical machines. When a small electromagnetic force is imposed on the electrical machines, the distribution of finite elements considerably affects the accuracy of electromagnetic force calculation. Although achieving symmetric distribution of finite element around the target is an effective scheme, it is not applied to any cases. To improve the calculation accuracy of the electromagnetic force, an error correction method based on the nodal force method is proposed herein. In the proposed method, although no special implementations are required, the finite element analyses are performed in the cases where the electromagnetic force does not act on the target. The performance of the proposed method is illustrated in the magnetostatic field using a simple problem.

This paper deals with a 3-phase, 36-slot per stator, and 24-pole double stator topology of an axial-field switched reluctance machine (AFSRM). A multi-slice automatic Mesh-Based Generated Reluctance Network (MBGRN) is used to model the machine. The automatic MBGRN model is used to achieve calculation of self-flux linkage and self-inductance per phase and evaluate the torque when the motor is fed with 3 phase step current. This topology has never been studied using multi-slice reluctance network modeling method. The developed multi-slice model allows fast simulation in 2D domain, still accounting for the inherent 3D effect of the axial-field switched reluctance machine. Validation of the presented modeling approach is made through comparison between the simulation results for the multi-slice meshed reluctance network model and a full 3D finite element model.

The accurate prediction of coupled inductive and capacitive effects in electromagnetic coils is becoming of crucial importance in several industrial applications, either due to the increase of operating frequency or to the increase of voltage levels. In this paper we propose to use 2D finite element formulations to identify resistive, inductive and capacitive effects at different modeling levels of electromagnetic coils, from turn-level to winding-level, and extract the corresponding lumped circuit parameters. The resulting models, whose computational complexity is compatible with current industrial design practice, aim to predict the first resonances as accurately as experimental characterizations, which is confirmed by simulations and tests performed on a typical high frequency transformer.

In this paper, a new hybrid excited axial electrodynamic wheel (HEAEDW) for EDS maglev is proposed and investigated in which propulsion and levitation forces are simultaneously created via moving magnetic rotor above a conductive guideway. The proposed machine is a hybrid excitation machine, which can not only realize control actively and thrust force, but also reduces the risk of PM demagnetization. The field and forces created by this machine are analyzed by using the analytical and finite-element method. Then, the parameters that affect performance were investigated and adjusted to improve performance. Finally, a HEAEDW prototype is manufactured, and experiments are carried out to verify the predicted results.The results show that the proposed machine is capable of producing not only an equivalent suspension force of the basic axial EDWs but also the thrust and controlled mode.

In this paper, a new methodology for the sensitivity analysis of the electric potential distribution near the transmission lines modeling by finite element method and using the adjoint method related to cable positions is proposed. This sensitivity of the objective function can be adopted during the optimization process by methods based on gradient information. The sensitivity of a large number of conductors by phase is obtained with high precision and very low computational cost using adjoint sensitivities, which is not dependent on the number of interesting design variables. The approximation using central finite differences to obtain sensitivity is adopted for comparisons and validation. With this methodology, the numeric model of the transmission lines can be adopted during the optimization process without compromising the required computational processing time.

This paper proposes a novel magnetic circuit design and size reduction for horizontal linear vibration motor (HLVM) used for vehicle LCD panels. The prototype of HLVM uses two thick permanent magnets to create a magnetic circuit sited below the voice coil, but the novel design places four thin permanent magnets above and below the voice coil. Meanwhile, the coil position changes to the yoke center to make a more effective magnetic circuit with shorter magnets. Compared to vertical linear vibration motor, the force calculation of HLVM is quite different. In this paper, a new force calculation method is used to approach the electromagnetic-mechanical coupling analysis of HLVM. The novel design has a smaller volume (-28.85%) but similar acceleration as a prototype. Experimental results verify the analysis results of HLVM (displacement, acceleration on dummy jig).

Permanent magnet type low-field and lightweight magnetic resonance imaging (MRI) system has the structure of iron yoke to provide closed magnetic paths. Ferromagnetic material with high permeability such as iron yokes can obviously influence the linearity of gradient field in the region of interest (ROI), leading to image distortion. The magnetic effect of ferromagnetic materials must be considered during the design process of gradient coil. This paper proposes a new design method for the gradient coil in permanent magnet type lightweight MRI system, as a supplement to the classical equivalent magnetic dipole method. The magnetization effect of the ferromagnetic materials is equivalently represented by the image magnetic dipoles, which are treated as the extra sources of the magnetic field in ROI except from gradient coils. The proposed method can effectively premeditate the iron effect during the design process. It is shown that the optimized gradient coils have good performance even under the influence of iron yoke.

As the torque characteristic in a surface permanent magnet synchronous motor (SPMSM) is strongly dependent on the magnetization distribution, the estimation of the magnetization distribution before the application of a permanent magnet to the SPMSM is essential to identified and exclude a defective permanent magnet. One of the fastest and most effective methods for identifying the magnetization distribution is based on the minimization of the square function composed of the measured magnetic flux density at the measured point, and the flux density derived from the Biot-Savart law. However, because there are countless magnetization distributions that can reconstruct the unique distribution of magnetic flux density on the outside of a permanent magnet, the achievement of realistic magnetization distribution is not assured. In this paper, a method for estimating a realistic magnetization distribution is proposed. The proposed method is composed of the Biot-Savart law and the mathematical programming. Furthermore, the sigmoid function is applied to all variables in the optimization problem to suppress the generation of unrealistic magnetization. The performance of the proposed method was compared with various estimation methods in the various permanent magnets.