Conference Agenda

It is well-known that real-world topology optimisation problems often exhibit multiple local minima and that derivative-based methods are prone to getting stuck in a possibly suboptimal local solution. We present a way of computing multiple locally optimal solutions to topology optimisation and illustrate the method for a two-pole synchronous reluctance machine.

Topology and parameter optimization of IPM motors can be used to design motor shapes. In this paper, an IPM motor with high torque and high mechanical strength is designed by incorporating stress and connection constraints in addition to electromagnetic field analysis. The multi-objective topology optimization was performed, and feasible Pareto solutions were obtained.

The present paper proposes a methodology based on topology optimization to design a synchro-reluctant motor considering magneto-elastic coupling. Stress on the rotor related to inertia forces and press-fitting effect on the stator are considered. The magneto-mechanical coupling is taken into account using a simplified magneto-elastic analytical model for the material behavior law. The objective is to maximize the machine's torque while considering an equality constraint on the volume. The topology optimization approach is based on a discrete BESO algorithm.

This paper proposes a generalized multi-material filtering formalism in density-based topology optimization. This work’s novelty is the consideration of the periodic and anti-periodic boundary conditions commonly used to simulate electrical machines, which change the nature of the materials located outside the simulation zone. It affects the filtering, which relies on a convolutive averaging of the materials’ properties near the boundaries. The implementation is detailed and applied to the topology optimization of a permanent magnet machine rotor.

We investigate a PDE constrained design optimization problem with an uncertain
parameter. By utilizing a robust worst case formulation we obtain an optimization problem of
bi-level structure. To obtain tractability we approximate the worst case function with a linear
Taylor expansion. Numerical results are presented for validation.

In the present work we propose a different approach to simulate and optimize a
rotating electric motor. We solve the eddy-current equation, with the use of the shape derivative
and space-time finite element methods. We then also consider electro-thermal coupling.

This presentation considers the problems of designing passive devices for mode conversion and multiplexing in the context of acoustic and electromagnetic wave propagation using topology optimization. Although these problems may appear similar, this talk focuses on the differences between the two cases and, in particular, the unique features of the electromagnetic problem.

This work presents topological optimization method to design magnetic cricuit in 3D. This method is based on adjoint method using the equations of volume integral methods. Thus, our approach makes it possible to reduce the computational time and the memory compared to classical topology optimization methods based on finite element calculations.

In this paper, density-based topological optimization is used to design a simple magnetic circuit. This study aims to minimize the mechanical compliance of a circuit subjected to a surface force and under a constraint in order to generate a specific magnetic field in a target zone. We propose an approach based on the SIMP approach and the adjoint method to solve the topological optimization problem applied to design 3D magnetic circuit.

This work proposes a method to perform multi-objective optimization of inductors using a surrogate model based on neural network (NN) instead of finite element method (FEM). Traditional design methods require repetitive FEM calculations, resulting in a very long overall design time. In contrast, this work utilizes a well-trained neural network to predict magnetic core loss, the volume and saturation current, avoiding repetitive FEM evaluations and significantly reducing the total time for inductor optimization.

The possibility of adopting data-driven procedure to create a model of electromagnetic problems has been investigated since long. Recently, the availability of high-performance computing systems even at desktop level has provided different model of “artificial intelligence” processors able to deal with such a problem, examples being Physically-Informed Neural Networks, Generative Adversarial Networks. Etc. In this study, using a simple yet representative benchmark problem, some of the most common solutions are compared, with the aim of highlighting respective advantages and drawbacks.

This article propose an optimal design of control gain parameters using a Cauer ladder network with constant basis functions and an Impulse-response model based V-Tiger. We model the TEAM Workshop Problem 28 by the Cauer ladder network method, which is a model order reduction technique, and enable the analysis by a circuit simulator such as Simulink. We determine the control gain parameters by applying the Impulse-response model based V-Tiger to the data obtained from the analysis. This method is useful for efficiently determining control parameters on simulation.

This paper discusses a novel technique for quasi-global parameter tuning of high-frequency structures using response features and inverse surrogate models. Our approach enables a low-cost identification of the most promising parameter space regions, followed by a fine tuning by means of local routines. The presented method is validated using several high-frequency structures. Its global search capability and computational efficiency are demonstrated by extensive comparisons with multiple-start local search as well as nature-inspired optimizers.

This paper presents a novel multi-material topology optimization method for multi-segmented permanent magnet motors. In the proposed method, optimization process is divided into two stages. In the first step, multi-material topology optimization, in which magnetization vector of each magnet element is defined stochastically, is performed to determine the rough magnet arrangement. In the second step, the geometry obtained in the first step is used as the initial solution, and detailed design is performed. To validate the effectiveness, the proposed method is applied to the shape optimization problem of a multi-segmented permanent magnet motor.

This paper presents a topology optimization of on-chip planar inductors based on evolutional on/off method and CMA-ES. The conductor shape of the inductor is expressed through the spatially-smooth function and its variables are optimized by CMA-ES. It is shown that the resultant inductor shape is varied depending on the given specifications.

In this paper, topology optimization and the frozen permeability method is used to verify the effect of asymmetric flux barriers on torque ripple reduction. Although an interior permanent magnet motor with an asymmetric flux barrier has been proposed in previous studies, the number of references is limited and the mechanism is unclear. In this paper, the geometry of symmetric and asymmetric flux barriers is optimized using topology optimization. Torque analysis is then performed for the obtained optimized geometry, and the frozen permeability method is used to separate the cogging torque and the load torque. The torque separation results reveal the torque ripple reduction effect due to offsetting of cogging and load torque, which has not been reported in previous studies.

We derive surface integrals for calculating the Lorentz force acting on a non-magnetic
conductive specimen when moving in the field of a spherical permanent magnet. The integrals
are valid for any geometry of the specimen, moving direction, position, and orientation of the
magnet. We evaluate the performance of this approach on a thin and thick cuboid, thin disc,
sphere, and a thin cuboid containing a surface defect. The normalized root mean square errors
are below 0.4% with respect to a reference finite-element solution.

The reconstruction of active sources in the human brain based on electroencephalography (EEG) is challenging in many respects. For high-quality clinical source reconstruction good signal quality, individual head images for realistic volume conductor models, as well as complex software tools for data processing are needed. However, there is an increasing demand for mobile EEG applications and, thus, the need for source reconstruction in areas like sports science or rehabilitation. In this study, we explored the possibility of using dry EEG electrodes for source reconstruction applying an averaged head model for forward computation, and developed an easy-to-use MS-Windows-based data processing pipeline in Python. Right-hand motor execution (ME) and imagination (MI) EEG data, which are typically used in brain-computer interface scenarios, were used to test our pipeline. Preliminary results show enhanced source activity over the contralateral motor cortex for both, ME and MI. We could demonstrate that source localization is feasible for mobile dry EEG scenarios such as in rehabilitation applications.

The magnetization of commercial permanent magnets is uneven, and the spatial magnetic field is distorted by individual differences. Therefore, it is necessary to measure the magnetic field of each magnet, but measuring the three-dimensional magnetic field distribution takes time. Thus, we measured the magnetic field near the permanent magnet and estimated the spatial magnetic field using two methods: the Gaussian process regression method and the truncated singular value decomposition method. We also calculated the average error for the measured values in two methods. As a result, we found that the average error of the truncated singular value decomposition method was smaller than that of the Gaussian process regression method.

The main goal of this paper is to analyse the influence of the shape and duration of the excitation pulse on the quality of reconstructed images in the inverse problem of the magnetoacoustic tomography with magnetic induction (MAT-MI). To solve this problem, the obtained reconstruction images were subjected to a binarization process, and their quality in relation to the original image was determined using a correlation and PSNR indicators. The research was conducted based on the reconstruction results obtained for several different shapes and durations of the excitation pulse.

In accelerator systems, magnetic field analysis incorporating the effect of magnetic hysteresis is essential for achieving high efficiency in beam commissioning using electromagnets. The play model is one of the hysteresis models that uses a shape function generated from measured BH loops as input. However, numerous BH loops are needed to account for minor loops, such as those caused by beam misalignment of accelerator magnets, and these measurements are time-consuming. Therefore, we aimed to develop a beam commissioning method that utilizes the play model by interpolating arbitrary hysteresis loops from measured data using Gaussian process regression in combination with the reduced play model which we have already proposed.

This paper presents novel procedure for optimizing the shape of conductive shields for low and medium frequency magnetic fields using convolutional neural networks (CNNs). In general, shape optimization is a specific class of so called ill-posed inverse problems, because objectives have typically many local minima when varying shapes. In the first approach a CNN is used as a surrogate model of the forward problem, while in the second approach a CNN is trained to solve the inverse problem directly. The case study presented in this paper involves the shape design of an electromagnetic shield placed in a sinusoidal uniform magnetic field under given constraints.

This paper first presents analytical expressions for the effectiveness of shielding time-harmonic magnetic fields by means of a rotating double-shell cylindrical shield made of non-magnetic material. Next, the procedure is described for finding the optimal distance between the two shells and their thickness (the same for both shells) for the maximum reduction of the weight of the double-shell shield (which is equivalent to reducing the active cross-sectional area) compared to a thick single shield for a given value of the shielding factor.

Magnetic particle-based targeted drug delivery is gaining momentum in recent years.
Significant challenges remain when it comes to modelling the movement of particles and achieving
precise targeting to desired regions. In this study, we address these challenges by training a
data-driven model to accurately predict particle velocity from magnetic particle positions and
electromagnet currents in an in-vitro targeting setup. The model is successfully applied to a
control algorithm to actuate particles from an initial to a predefined final position.

In recent years, the development of electric vehicles is being actively pursued. Several permanent magnet synchronous motors are installed in xEVs. To improve the manufacturing quality of PM motors, it is essential for revealing the magnetization state inside PMs at the early design stage. Therefore, a method called SiGrad has been proposed to numerically estimate the magnetization distribution using the magnetic flux density measured around PMs. As a result, the effective performance of SiGrad is illustrated in the estimation problem of PM with the homogeneous magnetization. However, its performance regarding the estimation of inhomogeneous distribution with demagnetized region is not investigated. In this paper, the applicability of SiGrad to the demagnetized magnetization estimation is verifed, and Pinching-type SiGrad (P-Sigrad) is proposed to enhance the estimation accuracy of SiGrad.

In recent years, because the carbon neutral evokes the widespread of electrification, the demand of electric vehicles (xEV) is increasingly growing. The DC-DC converter loaded on xEV generates electromagnetic noise (EM) that can interfere with other equipment. Therefore, an effective electromagnetic shielding around the wire harness connected to the DC-DC converter is required by the design optimization method. In this paper, to establish the design method based on the topology optimization, the accuracy of the sensitivity analysis using the time domain adjoint variable method is investigated in the electromagnetic shielding model.

The reconstruction of the spatial complex conductivity σ + jωε0εr from complex valued impedance measurements forms the inverse problem of complex electrical impedance tomography, or complex electrical capacitance tomography, respectively. Regularized Gauß-Newton schemes have been proposed for their solution. However, the necessary computation of the Jacobian is known to be computationally expensive, as standard techniques such as adjoint field methods require additional simulations. In this work we show a more efficient way to computationally access the Jacobian matrix. In particular the presented techniques do not require additional simulations, making the use of the Jacobian free of additional computational costs.

An automatic design optimization of a wireless power transfer system is performed using Monte Carlo tree search. Several key factors, i.e., the compensation network, shapes and geometrical parameters of the core are determined after searches, in order to achieve the high transfer efficiencies for nonmisaligned and misaligned cases. This work demonstrates the effectiveness of Monte Carlo tree search in achieving design optimization for wireless power transfer systems.

The eddy current loss of stator windings in motor due to the skin and proximity effect cannot be ignored, especially when the input carrier harmonics increases. In this work, a semi-analytical homogenization method is used to calculate the impedance of a single tooth model in permanent-magnet synchronous motor. An optimization problem is solved to identify the circuit parameters in the Cauer circuit which has the impedance obtained by the homogenized analysis. The Cauer circuit can be used for a more effective time-domain analysis considering magnetic hysteresis.

This study proposes an optimization method for magnetic shielding in eddy current systems through the design of conductive shields using the level set method. The shape of the conductive shield was represented through the level set function, and the shape deformation was performed by solving the level set equation. The expression of the design velocity in the level set equation was defined based on the continuum sensitivity of the eddy current system to the conductive shield shape. The proposed method was verified by applying it to the conductive shielding design of the magnetic induction tomography system.

The design and optimization of the electromagnetic layout of E-Machines in electric vehicles play a crucial role in achieving high-performance and efficient propulsion systems. This abstract paper explores the significance of optimization techniques in the layout of electromagnetic components within electric vehicle (EV) powertrains, with a specific emphasis on addressing noise, vibration, and harshness (NVH) considerations.

Various aspects of the electromagnetic layout optimization are discussed, encompassing the geometric arrangement of motor components such as stator winding, rotor structure, and permanent magnets, as well as the selection of suitable materials. The paper addresses the challenges associated with striking a balance between compact design, cooling efficiency, electromagnetic performance, NVH characteristics and cost . An emphasis is placed on conducting multi-objective optimizations to obtain optimal E-Machines that excel in performance and NVH metrics.

The utilization of advanced simulation tools and optimization algorithms is explored, providing engineers with the means to model and analyze electromagnetic characteristics, evaluate design alternatives, and identify areas for improvement. The paper highlights the benefits of simulation-based optimization, which include reduced development time, cost savings, and enhanced design accuracy.

Furthermore, the paper presents case studies and real-world examples where optimization techniques have been successfully applied to the electromagnetics layout of E-Machines in electric vehicles. These examples illustrate the practical implementation of optimization methods and their impact on improving motor efficiency, reducing losses, and enhancing overall vehicle performance while considering NVH aspects.

Sensitivity analysis (SA) is performed in this work, including various parameters and
considering different transmitting (TX) antenna types in reverberation chambers (RCs). To this
end, surrogate modeling techniques were involved to efficiently calculate the Sobol’ indices as a
measure of uncertainty quantification (UQ). This approach helps to appraise the contributions
of different parameters in the lower frequency range, where the well-stirred condition cannot be
established, yielding a proficient apparatus for the stirrer and the chamber design.

Parasitic electromagnetic (EM) effects within passive components are a fundamental issue in EMI filter applications when certain filter specifications must be met. In this work, a surrogate model based optimization methodology is used to identify an ideal geometry of a surface-mounted (SMT) ferrite bead to reduce its parasitic capacitance while a maximal inductance and consequently a maximal impedance is provided. The identified geometry exhibits an ideal device behavior up to 1 GHz.

In this work, two different approaches for solving an inverse problem to determine the local magnetic material properties of electrical steel sheets are compared. The first approach involves a quasi Newton method approximating the Jacobian with Broyden's update formula and the second is an adjoint method. To handle the ill-posedness of the inverse problem, a Thikonov regularization is used for both methods and the regularization parameter is computed via Morozov's discrepancy principle.

The aim of this work is to optimize the sensor positions of a sensor-actuator measurement system for identifying local variations in the magnetic permeability of cut steel sheets. Before solving the actual identification problem, i.e. finding the material distribution, the sensor placement of the measurement setup as well as the positions of the system relative to the steel sheets should be improved in order to increase the identifiability of the material distribution. For the objective function of this design optimization the Fisher information matrix (FIM) is used, which allows to quantify the amount of information that the measurements carry about the unknown parameters. The forward problem is solved by the finite element method.