Poster Papers

In the design process of an accelerator cavity, optimization of the geometry parameters is necessary in order to tune the device. Finding the ideal set of parameters can be challenging and numerically expensive, as each evaluation requires the numerical solution of Maxwell’s eigenvalue problem on the geometry at hand. The goal of this work is to efficiently employ a gradient-descent based optimization approach which takes multiple tuning objectives into account and ensures a correct matching of modes.

The B(H) curve of yoke laminations may be an uncertain parameter in numerical field models of accelerators due to manufacturing variances or because sample magnetometer measurements may not be available. In this work, a method is derived to compute the B(H) curve of iron yokes, for a given set of material measurements on ring specimens and field measurements of the built magnet. Instead of using a closed-form expression for the B(H) curve, material measurements are used to derive a data-driven model that reflects the observed variances due to chemical compositions, heat treatment and cold working. To this end we make use of a truncated Karhunen-Loeve expansion. The parameters of the data-driven model are subsequently updated by fitting the simulated to the measured magnetic flux density in the magnet. It is shown that the proposed method can retrieve a previously selected ground truth B(H) curve that was used to generate the field data for the fitting.

In this work we propose a new Hermite least squares optimization method for problems in electrical engineering. The aim is to solve bound constrained non-linear optimization problems, where the derivatives of the objective function are available with respect to some optimization variables, while for others they are not. This method is highly relevant for failure probability minimization having deterministic and uncertain optimization variables.

Magneto-static finite element (FE) simulations make numerical optimization of electrical machines very time-consuming and computationally intensive during the design stage. In this paper, we present the application of a hybrid data-and physics-driven model for numerical optimization of permanent magnet synchronous machines (PMSM). Following the data-driven supervised training, deep neural network (DNN) will act as a meta-model to characterize the electromagnetic behavior of PMSM by predicting intermediate FE measures. These intermediate measures are then post-processed with various physical models to compute the required key performance indicators (KPIs), e.g., torque, shaft power, and material costs. We perform multi-objective optimization with both classical FE and a hybrid approach using a nature-inspired evolutionary algorithm. We show quantitatively that the hybrid approach maintains the quality of Pareto results better or close to conventional FE simulation-based optimization while being computationally very cheap.

In this paper, a comparison between the model order reduction technique and deep learning technique is proposed, from the aspect of the surrogate model construction in computational electromagnetism. The merit and demerit of both approaches are discussed and compared via an academic application of magneto-thermal coupled analysis.

First results of a research project aiming to identify locally the magnetic properties of cutted electric steel sheets are presented. In doing so, we provide the idea of a sensor-actuator system being able to excite and measure locally the magnetic field, and an inverse scheme based on solving the magnetic partial differential equations with the finite element (FE) method. First results are discussed, in which the measured data is obtained by a forward simulation and we restrict to locally varying linear magnetic permeabilities.

In this paper, the distribution of electromagnetic from the two-dimensional (2-D) finite element analysis of the simplified transformer is estimated by the convolutional neural network (CNN) model U-net improved with ResNet. By changing the geometric dimension, material property and current excitation, the image dataset with different parameters is obtained and additionally expanded by scaling, flipping and rotating. Based on it, the weight parameters of improved U-net model are trained and the optimization for the model is carried out by hyperparameters searching, which improves the estimation precision of the electromagnetic field distribution of the transformer. The estimation results prove the effectiveness of the method.

The accurate hysteresis characteristics modeling of ferromagnetic materials is crucial to optimally design of electro-magnetic equipment. In this paper, the hysteresis play model is improved to simulate magnetic characteristics of ferrite at variable temperature conditions based on recurrent neural network. The proposed model predicts the magnetic field strength of N87 ferrite with the Mean Square Error (MSE) better than 0.3%. The results demonstrate that ferrite materials can be accurately described by the improved model trained with a limited amount of data.

This study introduces a topology optimization method with an explicit and geometrical way that can represent the shape of air area with vertices. By applying genetic algorithm to optimize the position of vertices, better torque property can be obtained. It is shown that the proposed method has a better performance compared to conventional methods in two aspects: better torque property and practical structure.

In this paper, the different hybrid optimization algorithms have been applied to the constrained optimization of the line-start permanent magnet synchronous motor. The optimal results for the hybrid Cuckoo Search and hybrid Grey Wolf Optimization have been compared. In the primary objective function, the parameters describing motor parameters in the steady-state have been included. Moreover, the non-linear constraint function has been taken into account. The optimization procedures have been elaborated in the Delphi and Matlab environments. The mathematical model of the line-start permanent magnet motor has been developed in Maxwell environment.

Manufacturing errors (ME) are ubiquitous and inevitable in product engineering manufacturing. However, the existing methods are hardly oriented to the discrete-variablebased modelling methodology for the topology optimization (TO). In this regard, a novel methodology based on morphological operations and random filed (MORF) is proposed for the discrete-variable-based topology optimization procedures to consider MEs. Morphological operations are firstly introduced to generate the geometrical variation. Moreover, the dimension of the structuring element in the morphological operations is set as the output of the random field function. Using the proposed approach, MORF is capable of quantifying spatially nonuniform MEs rigorously. The numerical result has validated the proposed method.

This paper presents a Reduced Order Model (ROM) workflow to calculate the radial and tangential forces required for Noise, Vibration and Harshness analysis on the stator teeth of an electrical machine over a selected operating range. The workflow is applied to a permanent magnet synchronous motor Finite Element (FE) model, and a Design of Experiments is used to generate the sample points to create the ROM via Response Surface Methodology. The results from the ROM and FE model are analysed and compared for consistency and accuracy, for different operating points.

This paper presents a multiresolution dynamic mode decomposition (MRDMD)-based approach to analyze electromagnetic near-field scanning. By dealing with the time-varying near-field radiation as a spatial-temporal correlated signal, MRDMD runs the conventional DMD in a recursive form. The fast and slow modes are distinguished according to the distribution of the derived DMD eigenvalues. The fast mode generates the input data for the next DMD execution. Finally, the multiresolution time-frequency representation of the near-field scanning data is derived. The corresponding spatial distributions of each frequency component are also obtained. The proposed MRDMD is benchmarked by a numerical example of switching on/off loop antennas. It is found that the obtained multiresolution timefrequency representation accurately shows the excitation duration and frequency components of each loop antenna, whose corresponding spatial distributions are also extracted correctly. This work provides a multiresolution analysis tool for near-field scanning, especially when the spatial-temporal information of different frequency components is required.

This paper presents an overview of the researches done at the Ampere lab on the use of time reversal since 2012. The first project aimed at optimizing the energy transfer in a complex propagation medium dedicated to a remote system wake-up. The proposed solutions consisted in the design of an optimized broadband rectenna and the use of specific transmitted signal waveforms. In addition to the optimization of the rectenna, the optimization of the transmitted signal has been addressed in order to exploit the properties of the channel. Time reversal is the optimized signal in this case. The second project aims at optimizing power and information transfer in the context of passive UHF RFID technology. With the constraints of exploiting existing commercial tags and respecting as closely as possible the standards in force, the objective is to explore the potential of RFID in pulse mode while exploiting the time reversal technique

This paper deals with the diagnosis of a fuel cell (FC) by magneto-tomography. The magnetic field generated by any electromagnetic device such as a FC can give important information about the health state of the FC. The magnetic field around the FC is directly connected to the current density distribution inside the FC. In this paper, the innovation consists in moving a ferromagnetic concentrator, composed of 2 Mu-metal elements separated by an air-gap where a Hall effect sensor is embedded, around the FC. This high-performance tool will enable to detect and locate more precisely a defect and, consequently, to establish a health state of the FC. The numerical analysis and comparison will validate the interest of the mobile magnetic field concentrator.

As wireless networks continue to advance and the need for low-latency communication links increases, calibrating antennas installed in the field is essential. In this context, Unmanned Aerial Vehicles (UAVs) are very useful for applications such as UAV-based measurements. Given its light weight, wide bandwidth makes Printed Log-Periodic antenna (PLPDA) an ideal solution as UAV probe. Our study examines the performance of a PLPDA mounted on a UAV. Extensive simulations are performed to determine the optimal position for the PLPDA on a UAV. Simulations are carried out in CST Studio Suite 2022 using the time domain Finite Integration Technique (FIT) with appropriate  mesh settings. At the optimized location PLPDA achieves a -10 dB bandwidth of 6.2 GHz.

The transient characteristics for three different constructions of electrodynamic railguns were simulated and compared with the measurement curves. They concerned the ironless (IL), iron-core (IC) and those iron with permanent magnet (ICPM) railguns. The rail dimensions and configurations as well as the supply systems were the same. The simulations were carried out using the own field-circuit numerical model. The modelling results were compared with those obtained from measuring tests and good agreement was obtained.

An influence of the capacitor configuration in the pulse supply system for rail accelerator on its efficiency was investigated in the paper. Two different configurations were compared: the first one with parallel connection of capacitors, and the second one with parallel-series connection. In both cases the same number of capacitors was assumed. All other parameters of the device were kept constant, i.e. mass and dimensions of the projectile, rail length and stator geometry. A field-circuit model was used to analyze numerically the transients for different capacitor configurations and voltage values.

Under the premise that the global is advocating low-carbon development, vegetable insulating oil has

attracted widespread attention as a green and clean material. Compared with traditional mineral insulating oil, vegetable insulating oil have the disadvantages of poor fluidity and insufficient insulating performance. Nano-doping modification is a commonly used insulating oil modification method. The vegetable insulating oil models with different doping mass fractions of nano-silicon dioxide, nano-aluminium oxide, nano-silicon dioxide modified with silane coupling agent (KH550) and nano-aluminium oxide modified with silane coupling agent (KH550) were established using Materials Studio

simulation software to explore the optimal conditions for nano-modified vegetable insulating oil. By analyzing the non-bonded interaction energy between nanoparticles and vegetable insulating oil molecules, the Mean Square Displacement (MSD) and the DC breakdown voltage of the modified system, it was explored whether the modified vegetable insulating oil could stably exist and the modified conditions with the best heat dissipation and insulation characteristics. Finally, the optimal conditions for nano-doping modified vegetable insulating oil were comprehensively analyzed, including the type and doping concentration of nanoparticles.

In an accelerator system, it is necessary to adjust beam orbits by changing the output magnetic fields with the current values exciting accelerator magnets. However, the magnetic hysteresis is not considered in this beam commissioning process. Instead, the current values are currently modified by trial-and-error. Hence, it is required to establish a prediction method of the current values that reproduce the magnetic fields including the magnetic hysteresis effects. In this paper, the reduced play model (RPM) is proposed. The RPM is constructed by re-identifying the shape functions from the I-B characteristics obtained from a conventional finite element method (FEM) incorporating the dc hysteresis effects. As a result of the examination, it was verified that the proposed RPM speeds up the magnetic field calculations which is fast enough for the usage during the beam commissioning.

Generally, bondwire is main radiation point at chip level emi. Recently, EMI radiation is seen in the DIE region of the LPDDR5 center-placed pad DRAM. While changing the DIE RDL structure, EMI simulation was performed using SIwave to determine the cause of EMI radiation. By dividing into an ideal structure and a structure that increases radiation, it is possible to identify the cause of EMI occurrence and effectively simulate EMI. It was confirmed that EMI radiation by the center-placed pad occurs when the current flowing through the power and VSS is disturbed by the slit/slot. Using this, it is possible to perform a more sophisticated EMI simulation

We design shim coils to homogenize the magnetic field generated by realistic Helmholtz coils utilizing the Truncated Singular Value Decomposition (TSVD) method. The optimized shim coils shape depends on the selection of the evaluation surface. In this study, two sets of different evaluation points are investigated to design shim coils.

Numerical and experimental studies on structural control of multi-layered high-temperature superconductor tapes at room-temperature using eddy-current thermography are presented. A rotating magnet wheel is used as an inductor. A three-dimensional integro-differential formulation in terms of the electric vector potential is developed for eddy current calculations in the HTS tape, where the integral part is discretized by a collocation method, and the differential part is discretized by the finite difference method, coupled with a thermal modeling using the diffusion equation, discretized by the finite difference method. The source magnetic field is calculated using the surface magnetic charge model. Simulation results are confirmed by measurements.

In this work, first results of the splitting of hysteresis measurements, performed with a rotational single sheet tester (RSST) into a reversible and an irreversible magnetic field component, are presented. This is of utmost importance for identifying parameters of energy-based vector hysteresis models because in their original formulation it is not possible to correctly depict vanishing rotational losses in full saturation. With the help of the proposed splitting of the measured magnetic field strength, new adaptions of the material model can be designed, which actually depict the underlying physical phenomena instead of just using a mathematical construct.

The paper deals with research on the direct and quadrature axis cross saturation effect in the synchronous reluctance machine (SynRM). The comparative analysis between performance of the 4-pole SynRM of 3-phase and 6-phase windings have been conducted exploiting developed 2D finite element models. In the numerical models of studied machines, the locked rotor method was adopted in which the direct and quadrature axes currents Iq Id were forced to determine magnetic fluxes in the d q axes, respectively. Determined magnetic flux maps have been used next to determine simple analytical model taking into account the cross saturation effect. The quality of proposed analytical description of cross saturation phenomena has been evaluated by comparison to the results of simulations in order to assess the possibility of adapting this description in the control algorithms of SynRM drives.

In this study, electromagnetic analysis of two UHF communication antennas integrated on F-4 aircraft is performed to improve the operational performance of these antennas by arranging their placements [1]. In parallel with this, in order to see the influence of main structures (such as fuselage, nose, wings and tail) composing the F-4 aircraft on the electromagnetic coupling between two UHF antennas, analysis is carried out on three different F-4 models each of which is modified such that main structures composing the aircraft are removed one by one. Finally, for the optimal antenna placement, far-field radiation pattern performance of the lower UHF antenna is analyzed in terms of directivity and coverage to see whether that antenna has sufficient directivity and coverage values. Aircraft and antennas are modeled based on their real electrical and physical properties in Computer Simulation Technology – Microwave Studio (CST- MWS) simulation tool.

Formulation [1] valid for magnetic and conductive wires of circular cross-section is extended to the case of rectangular cross-section. Comparison with FEM shows that the approximate approach guarantees a good accuracy.

This article presents a hybrid dynamic mode decomposition (HDMD)-based approach for the joint azimuth and elevation angles estimation via the uniform rectangular array. The modified and augmented DMD are hybrid in the proposed scheme. The received 2D spatial correlated signal is first decomposed through the modified DMD in terms of spatial DMD eigenvalues and corresponding eigenvectors. The imaginary part of spatial DMD eigenvalues represents the y direction’s electrical angle. Then, the augmented DMD examines the eigenvectors, which yield the x direction’s electrical angle. Finally, the azimuth and elevation angles are calculated from the electrical angles. The numerical example is conducted to validate the proposed HDMD. It is found that the azimuth and elevation angles are estimated accurately with automatic pairing. Our work offers a practical tool for source location in the MIMO systems.

In online single-sided partial discharge (PD) location settings, PD reflection patterns are affected by all components present in the cable circuit. This paper describes the performance of electromagnetic time reversal (EMTR) when interfering reflections contribute to the transient waveforms emitted by the PD. The analysed situation refers to a ring main unit (RMU) in the medium voltage (MV) grid where PD recordings are disturbed by signals reflected from the other cables connected to the RMU, potentially affecting the PD location accuracy. We show that the accuracy of EMTR-based location methods is unaffected by such effects.

We present a novel time-reversal spatial convergence metric, inspired by probability theory and the electromagnetic wave potential energy density, and apply it to several FDTD time-reversal simulations based on a modified Meep implementation. The proposed metric is a convex-quadratic function of time, attaining its minimum at a time and a value corresponding to the source power expected value and standard deviation.

Cauer Ladder Network (CLN) method is more and more used for order reduction of large numerical magnetoquasistatic model. It appears during the process of construction of the reduced model, loss of orthogonality of the vectors of the reduced bases which can lead to increase the error of reduction. To overcome this issue, a modified Gram-Schmidt process is introduced. The modified process of CLN construction is evaluated on a 3D magnetoquasistatic example in the frequency domain.

This paper discusses the effect of the location of measurement units on the electromagnetic time reversal (EMTR) based fault location method. It is shown that the location of the measurement point can greatly affect the accuracy of two EMTR based techniques (L2 and Lmax). Furthermore, the impact of the fault impedance has been studied. The findings are presented by considering the distance from the guessed fault location to the true fault location for a number of fault cases.

The Lanczos algorithm is applied to derive the Cauer ladder network for induction heating analysis. The equivalent circuit obtained by the first Lanczos algorithm is a Cauer-I type circuit consisting of thermal inductance and resistance. Reapplying the Lanczos algorithm to the Cauer-I type circuit equation yields a Cauer-II type circuit consisting of a thermal capacitor and resistance, as in a typical thermal circuit. The temperature distribution can be reconstructed by superpositions of the basic  functions corresponding to the circuit elements.

This paper generalizes the electromagnetic time reversal (EMTR) in mismatched media based bounded phase property in locating faults in multiconductor transmission lines. The so-called direct-reversed-time transfer function is first derived in matrix notation as an analogue of its original expression in two-conductor transmission lines. The bounded phase property is then extended in both symmetrical and unsymmetrical fault scenarios.

We present an electromagnetic Time Reversal MUltiple SIgnal Classification (TR-MUSIC) method to localize landmines. The performance of TR-MUSIC is investigated using numerical simulations of a two-dimensional configuration considering the effect of noise. The performance of the TR-MUSIC is also evaluated experimentally in free-space.

Numerical simulations are used to determine electric fields induced by low-frequency magneto-quasistatic fields into exposed human bodies. One of the utilized numerical methods is the Co-Simulation Scalar-Potential Finite Difference scheme that requires the magnetic flux density distribution, the field frequency, and the electrical conductivities of the body tissues as inputs. All of these quantities are subject to errors. New advances in the implementation and execution time of this numerical simulation method on Graphics Processing Units allow to calculate the body-internal electric field strength within seconds. One million simulations require 11.3 hours on 40 NVIDIA A100 GPUs, offering the option to use a Monte-Carlo approach to determine the uncertainty of the body-internal electric field strength.

A framework for stabilizing the low-frequency instability that arises in electro-quasistatic field simulations, under the presence of non-conducting material, is developed. The resulting symmetric formulations rely on penalization for imposing the electric Gauß law in void, and hence, they constitute approximations of continuously extended problems. Real and imaginary penalty-weights  are studied numerically, in terms of accuracy and conditioning.

The paper deals with the finite element (FE) analysis of the temperature influence on functional parameters of the line start permanent magnet synchronous motor (LSPMSM). The two-dimensional (2D) FE model of coupled electromagnetic and thermal phenomena in the LSPMSM was described. The nonlinearity of the magnetic circuit and the influence of temperature on the magnetic properties of permanent magnets as well as on the electric and thermal properties of the materials have been considered. The simulation results were validated by measurements of the prototype motor, showing satisfactory concordance.

The port-Hamiltonian (pH) framework gives rise to math-ematical models that preserve physical quantities, such as the energy, or maintain dissipation inequalities. Within multiphysics problems, submodels that appear as port-Hamiltonian systems (pHS) with input/output quantities that are coupled linearly, for many applications with skew-symmetric matrices, result in global pH systems. In this contribution, the suitability of the pH framework is demonstrated for systems of equations that result from mimetic discretizations of electromagnetism, such as the finite-integration technique (FIT), for network formulations, and for thermodynamic field formulations.

The resonance issue is regularly found in EMC engineering. Different experimental methods were proposed to alleviate this EMC issue. This paper describes an original method of electromagnetic (EM) cavity resonance effect reduction. The developed negative group delay (NGD) method is based transfer function equalization principle. The bandpass (BP) type NGD design as an unfamiliar engineering is analytically formulated. The proposed method feasibility is shown with 42×28×3.8 cm-size cavity simulation by using RLCseries network-based BP-NGD active circuit. By means of about -4 ns NGD value at 0.644 MHz centre frequency, resonance effect reduction with 1-dB flatness is discussed. The transient results highlighting delay reduction and signal integrity (SI) improvement with input and output cross correlation from 89% to 99% are presented.