Conference Agenda

In this paper, we consider fast iterative methods for {solving} a Helmholtz type problem in soft magnetic materials. The suggested methods are based on domain decomposition methods of Schwarz type. The convergence of these methods depend on the choice of transmission conditions. The proposed methods reduce computational resources like the time-to-solution and the memory footprint compared to sparse direct solvers applied to the whole domain at once, an approach which is often used in practice. The main contribution of this paper is the use of the non-overlapping Schwarz based domain decomposition methods for estimating apparent material parameters of Mn-Zn ferrite cores in a high-frequency regime. We employ the proposed method for solving the electromagnetic full-wave equation and compare the results against measurements and results obtained by a classical finite element solver.

We propose a data-driven modified nodal analysis circuit solver, which has the capability to treat elements for which only measurement data are available. The solution of the circuit problem is computed by minimizing the distance between circuit states that are compatible with Kirchhoff’s laws to circuit states defined by the measurement data. Hence, the inevitable need for a model representation is abolished, which avoids the introduction of modeling errors and uncertainties. Furthermore, the workflow is tremendously reduced, saving time and costs. An RC-circuit is considered as a first numerical example.

Metasurface-based reconfigurable intelligent surfaces (RIS) are able to actively control the reflection and refraction patterns on a surface to create a programmable and reconfigurable wireless propagation environment. In this work, a fast EFIE (electric field integral equation) formulation is developed to model metasurface-field interactions for RISs. The novelty of the formulation is that it utilizes characteristic basis functions (CBF) designed specifically for approximating surface current densities on metasurfaces, thereby significantly reducing the necessary number of unknowns and decreasing the solution time compared with employing the standard Method of Moments (MoM) technique with Rao-Wilton-Glisson surface basis functions. A numerical example is presented.

Commercial finite element method simulation programs, such as COMSOL Multiphysics, provide multiple simulation models for electromagnetic analysis of linear current systems. However, the conventional models may be deficient in analyzing a nonlinear current system, e.g., a non-insulated high-temperature superconducting magnet system. For instance, they are not able to consider not only a highly nonlinear resistivity of a superconductor but also a non-insulated magnet configuration in simulation. Therefore, here we report a fully-coupled simulation model to analyze nonlinear electromagnetic behaviors induced by a nonlinear current system. It may address the deficiency of conventional models by coupling H formulation, A formulation, and I-V formulation methods. Meanwhile, the validity of the fully-coupled simulation model is proven by an experimental study.

In this paper, a co-simulation model is presented to study the vibrational behavior of a coupled 2-D finite-element model of an induction machine with a mechanical radial ball bearing model. This method is applied to generate vibrations signals of a ball bearing mounted on the rotor shaft in a frequency domain in healthy and faulty states through magnetomechanical coupling. This multiphysical way of approaching the problem gives a new numerical application to generate vibration data for different types of faults in the machinery, and further to build a thorough dataset able to train a machine learning classifier for condition monitoring of ball bearing in different operating conditions. The current results show that several severities of ball bearing ring defects can be identified at constant speed for a 100 N radial magnetic force of the rotor.

This paper describes two methods to extract a manufacturable coil from the continuous stream-function distribution on a given meshed surface. The first method connects consecutive iso-contour lines of the stream-function distribution using a blending function, chosen by the user. The second method is based on a fitted B-spline surface on the stream-function distribution, that uses the normalized gradient. Both methods result in a connected curve following the stream-function distribution, which can be used to create a three-dimensional volumetric coil. The obtained coils are compared on resistance and flux density difference between the continuous stream-function distribution.

In this paper, a high-accuracy analysis method and a method to improve convergence using self-equivalent circuits were studied. Modified Nodal Analysis was applied to analyze the self-equivalent circuit more precisely. A nonlinear solution was obtained by reflecting the nonlinear characteristics of the magnetic material and using the numerical analysis method (Newton-Raphson method). The circuits of the stator and rotor were densely divided and each part was modeled as a magnetoresistance with nonlinear characteristics. By dividing the permanent magnet part, the analysis of the V-type model is also possible. The characteristic analysis of the PMSMG can be performed using the flux linkage obtained by solving the solution of the system of nonlinear equations.

This paper presents a comprehensive design optimization of inductors using Monte Carlo tree search. In contrast to conventional optimization, the proposed approach can simultaneously optimize the geometrical parameters as well as configuration such as inductor size, number of turns and core material. An inductor is identified by choosing a route from the root to a leaf at which shape optimization is performed. The covariance matrix adaptation evolution strategy is used for the shape optimization. It is shown that the proposed method obtains an optimal solution whose characteristics meet the requirement.

Sensitivity analysis is a methodology that can be used to assess how sensitive certain quantities of interest within the analyzed system react to model parameters uncertainties.

To calculate the sensitivities in a multi-parameter setting, the adjoint method is the method of choice. To obtain time-dependent sensitivity information, many adjoint problems need to be solved. In this paper, the adjoint problem simulation is accelerated using the parareal method which can significantly decrease the wall clock time. The approach is illustrated for a simple half-wave rectifier and a B6 bridge-motor supply network.

The computational costs to determine the eddy currents in nonlinear laminated iron cores could be reduced essentially using multiscale finite element methods. However, they are still too high for routine tasks in the design of electrical devices. Therefore, this paper investigates the feasibility of the multiscale finite element method with model order reduction using the discrete empirical interpolation method to facilitate the handling of a nonlinear problem. Numerical simulations show very satisfactory results.

The multiscale finite element method is an efficient technique to solve the eddy current problem in laminated materials without the need to resolve the thin iron sheets in the finite element mesh. It has been shown to give a highly accurate approximation of the classical finite element solution for a wide range of applications. However, in practice it is often necessary to have a reliable a-posteriori estimator of the error to either fulfill precision criteria or to use adaptive refinement. For the multiscale finite element methods such estimators have been presented for a restrictive range of applications, all of them using a frequency domain formulation. This work extends the theory of the estimator to a formulation in the time domain. It presents the mathematical background of the estimator and demonstrates its accuracy in a two dimensional model example.

Integral methods to solve eddy currents problems are widely used since they only need to mesh conducting regions of a given domain. However, their main drawback is that they give rise to a dense stiffness matrix.

In this paper we will compare the two main approaches to exploiting the properties of this dense matrix to reduce its memory occupation and the computational time: voxel-based approaches, accelerated by the Fast Fourier Transform (FFT), and tetrahedral grid approaches, that better approximate curved domains and can be accelerated with the Fast Multipole Method (FMM).

The aim of this paper is to compare the performance of voxel-based approaches, represented here by the VoxHenry software, and tetrahedral grid approaches, represented by the recently introduced VINCO framework, in terms of accuracy, memory occupation and computational time for different geometries.

The Darwin model for quasi-static electromagnetic field analysis is of particular interest for the analysis of power electronic devices such as power transformers and wireless power transfer systems. the Darwin model is an approach that neglects the second derivative term in time of the displacement current and considers the electrostatic field. This approximation allows conventional techniques for quasi-static magnetic field analysis, such as considering nonlinear magnetic material properties and coupled circuit analysis, to be appropriated, and capacitance effects can be included. Several methods have been proposed for frequency-domain analysis. This is because it is difficult to properly derived formulation of time domain analysis. We have already proposed a potential formulation in the frequency-domain analysis and succeeded in obtaining appropriate calculation results. In this paper, we propose a time-domain analysis of the low-frequency stabilization formulation in the finite element method. The proposed formulation is based on the Coulomb gauge condition, which defines no additional variables and redundant variables to improve the convergence characteristics. In addition, the system matrix can be made symmetric for solving by general iterative solvers.

In this report, the power transfer efficiency prediction about non-contact charger systems using each coil wire in the 85 kHz band was considered. Firstly, the coils using vinyl-insulated copper wire and litz wire were prepared as the sending and receiving coils and the power transfer efficiency for each coils were compared by experiment. Next, these coil constants such as the self- and mutual inductances, the AC resistances were calculated by finite element method (FEM). After FEM analysis, the equations derived from the equivalent circuit for the non-contact charger systems were calculated by Runge-Kutta method. Finally, the calculation results were validated through comparison with experimental results.

A novel variational formulation for high-order frequency-sensitivity analysis was proposed and verified in A-ϕ formulation of eddy-current system. The A-ϕ formulation of eddy-current system was differentiated with respect to frequency for deriving the frequency-sensitivity. The high-order formulation was sequentially obtained from the lower-order one. The state-variables of the proposed formulation are the frequency-sensitivity. The frequency-sensitivity analysis was performed by commercial FEM software because the variational approach is adopted to derive the sensitivity. The numerical example of an asymmetrical conductor with a hole was used to verify the theoretical validity and the numerical applicability of the proposed sensitivity formulation. The numerical result showed that the computational efforts are drastically reduced by the sensitivity analysis when the frequency-response is calculated.