Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Session Overview
PB-M4: Numerical Techniques
Wednesday, 17/Jul/2019:
10:50am - 12:40pm

Session Chair: Xikui Ma
Session Chair: Gerard Meunier
Location: Patio 44-55

Show help for 'Increase or decrease the abstract text size'

Investigation of the Coil Failure modes in the Electromagnetic Forming Process

Shantaram Dond1, Hitesh Choudhary2, Tanmay Kolge2, Archana Sharma2, G K Dey2

1Homi Bhabha National Institute, India; 2Bhabha Atomic Research Centre

Generation of a high pulsed magnetic field is a prime requirement in the electromagnetic forming (EMF) process. Failure of the electromagnetic coil generating a high pulsed magnetic field is an important issue and yet it is not much reported in the past. Due

to complex multi-physics and transient nature of the EMF process, it is difficult to diagnose the cause of the coil failure practically. In the performed electromagnetic tube expansion experiments, it is observed that the solenoid coil is withstood till 145 kA peak

current having 18 T peak magnetic flux density but failed when operated at higher current. In order to diagnose the coil failure, a 2D numerical model is developed using COMSOL. A sequential coupled approached is introduced in the numerical model to solve magnetic, mechanical and thermal physics. It is observed that at higher currents the temperature of the coil is well above the temperature limits of the supporting insulator. The von Mises stress acting on the coil supporting insulator is less than the insulator withstanding limits. Based on experimental observations and numerical model results, it is concluded that thermal failure is the primary cause that leads to mechanical and electrical failure of the coil.

Efficient power converter simulation using transformed PWM basis functions with field-circuit coupling

Andreas Pels1,2,3, Herbert De Gersem1,2, Ruth V. Sabariego3, Sebastian Schöps1,2

1Graduate School of Computational Engineering, Technische Universität Darmstadt, Germany; 2Institut für Theorie Elektromagnetischer Felder, Technische Universität Darmstadt, Germany; 3Department of Electrical Engineering, EnergyVille, KU Leuven, Belgium

The simulation of switch-mode power converters with conventional time discretization is computationally expensive since very small time steps are needed to appropriately account for steep transients occurring inside the converter. An efficient simulation is obtained using Multirate Partial Differential Equations, which allow for a separation of the different time scales. Replacing parts of the circuit by field models for a detailed analysis leads to large equation systems. The novel transformed PWM basis functions allow for a decoupling of the equation systems and thus an efficient simulation of the field-circuit coupled problems.

Classification of Radially Homogeneous Multi-Layered Spherical Targets from Resonant Scattering Signal Characteristics

M. Alper Selver1, Mustafa Secmen2, E. Yesim Zoral1

1Dokuz Eylul University; 2Yasar University

Object recognition using electromagnetic scattering signals is a significant problem found in different areas of application. In this respect, due to their special properties, spheres are one of the most widely studied geometries in the literature. On the other hand, the challenges associated with target classification by resonant scattering signals from multi-layer spheres has not been investigated in detail. For this purpose, time domain scattered signals are generated numerically from two multi-layer and two single layer radially symmetrical dielectric spheres. These signals are classified using a well-established time-domain feature-classifier strategy and the performance is observed. The reliability under low SNR conditions is also analyzed and reported. It is observed that the late-time resonant scattering characteristics of multi-layer spherical targets are significantly different than those of single layer ones and can be used for classification of targets having the same geometrical, but different dielectric properties. However, it is also observed that further advancements are necessary to reach the accuracy, specificity and sensitivity levels as high as single layer targets.

Rotating Machine Analyses with Cylindrical and Flat Slide Surface using Non-conforming Mesh Method

Hirokatsu Katagiri, Akira Ahagon, Yohei Watanabe, Noriyo Mimura, Kensuke Matsunaga, Kazuki Semba, Takashi Yamada

JSOL Corporation, Japan

In the finite element analysis of axial gap motors and motor case stray load losses, the surface shape of the air gap is a mixed shape of cylindrical and flat surfaces that connect the rotating part and non-rotating part. In this case, non-conforming mesh is used in slide surfaces to eliminate mesh generation at every step. Non-conforming mesh is applied to cylindering slide surface or flat surfaces. It is conversely not applied to slide surfaces with a mixed shape of cylindrical and flat surfaces.

In this paper, non-conforming mesh is applied to slide surfaces with a mixed shape of cylindrical and flat surfaces. The limitations of this method are clarified. In this digest, we show the ICCG convergence and calculation time of the non-conforming method in parallel computing with the localized ICCG method as one of the limitations of the non-conforming method. In the results, the degradation of convergence with the localized ICCG method is the same for the non-conforming mesh method as well as the conforming mesh method. In the non-conforming mesh method, a 3.8 fold increase in speed is achieved in the calculation of the axial gap motor via parallel computing with 128 processes compared to the conforming mesh method due to the reduction of time for generating mesh and reading mesh data.

Separated Representation of the FE solution of the Nonlinear Magnetostatic Problem based on Non-Intrusive PGD

Thomas Henneron1, Guillaume Caron1, Stéphane Clénet2

1université Lille / L2EP, France; 2Arts et Métiers ParisTech / L2EP, France

The Proper Generalized Decomposition (PGD) has shown its efficiency to solve parameterized problems in engineering. In order to develop a non-intrusive implementation, an approach consists in evaluating the Finite Element (FE) model for different values of parameters and to seek a PGD approximation of the solution. Then, the PGD approximation

which is very fast to be evaluated, could be used for real-time applications. In this communication, this PGD approach is performed for a nonlinear magnetostatic problem and is applied with a single phase transformer.

Large-Scale Parallel Electromagnetic Field Analysis Based on Additive Schwarz Preconditioner

Hiroyuki Kaimori, Hassan Ebrahimi, Akihisa Kameari

Science Solutions International Laboratory, Inc., Japan

We apply the additive Schwarz method to large-scale parallel electromagnetic analysis. Additive Schwarz method is a domain decomposition method based on partitioning global solution domain into a number of subdomains, coupled through the domain interfaces. Interface solutions are solved as independent variables in neighboring subdomains. The matrices are solved independently, and the coupling is treated by the additive Schwarz preconditioner. In this paper we investigate the effectiveness and performance of the method in parallel electromagnetic field analysis. In addition, we apply the method to rotating machines with nonconformal mesh connection between a stator and a rotor meshes.

An Adaptive Refinement Method for Solving Ion-flow Field of HVDC Transmission Line

QiWen Cheng, Jun Zou, Jiansheng Yuan

Tsinghua University, China, People's Republic of

To improve the accuracy while solving the ion-flow field problem, an adaptive refinement method based on the upwind technique is proposed. By limited the residual indicator in the subdomain of each node, the numerical error can be reduced. Numerical example verifies the significant effect. This adaptive refinement strategy can depict more details of HVDC ion density.

AC Resistance Numerical Calculation of Radio-frequency Coil for Nuclear Magnetic Resonance Logging

Zheng Xu1, Xiaohan Kong1, Pan Guo2

1School of Electrical Engineering, Chongqing University,State Key Laboratory of Power Transmission Equipment & System Security and New Technology, 400044, China; 2College of Physics and Electronic Engineering,Chongqing Normal University, No. 12 Tianchen road, Shapingba, Chongqing, China

Nuclear magnetic resonance logging (NMRL) is an effective and widely-used petroleum logging method.The logging sensor of NMRL works in the cylindrical drilling cavity, to ensure effective utilization of space, the radio-frequency (RF) coil is attached to the outer side of the cylindrical probe and adopts curved structure. The curved radio-frequency coil (CRFC) is used to emit radio-frequency magnetic field and receive detection signal, accurate calculation of AC resistance of the CRFC is essential to optimize RF coil for maximum signal-to-noise ratio (SNR). NMRL systems work in the kHz- MHz frequency range. The current is non-uniformly distributed inside the conductor because of skin effect and proximity effect under high frequency conditions, furthermore, the casing for mud going through was made of stainless steel for strength and is close to the coil, the eddy current induced by the RF coils in the casing will in turn influence the current distribution inside the RF coil. And the ultra-thin(30-70um) structure of coil making field distribution inside the conductor extremely asymmetric. So the analytical calculation of AC resistance of curved RF coils is difficult and there is no closed-form solution. Finite-element method (FEM) is in common use to solve this problem but it requires a lot more computing time because the coil’s thickness is quiet small comparing with the scale of length and the skin depth is extremely small at high frequencies. An efficient non-equidistant mesh volume filament (NMVF) model is used to calculate AC resistance of CRFC with the presence of the adjacent metal cylindrical casing at high frequencies. With this method, both RF coil and metal casing are partitioned into filament loops with constant current densities, transferring the electromagnetic field coupling effects to the circuit domain. This method can evidently reduce computing time compared with FEM. And the validity of this method are verified by measurement and FEM results.

Identification of Electric Multipolar Sources in Presence of Different Conducting Layers

Olivier Pinaud1, Laure-Line Rouve1, Olivier Chadebec1, Cedric Goeau2, Arnaud Guibert2

1Univ. Grenoble Alpes, CNRS, Grenoble INP, G2Elab, France; 2DGA, France

This paper deals with the multipole expansion identification of complex electric sources generating a static current flowing in a conductive medium bounded by layers with different conductivities. The expansion series in homogeneous medium is enriched by additional terms resulting from the generalization of the image method applied to multipoles. From close electric measurements, the solving of an inverse problem allows determining the multipolar sources, whatever the conductive environment. The efficiency of the approach is demonstrated through a numerical test.

Simple method to reduce computation time in planar airgap 3D FEM non-linear problems.

Dominique Giraud1,2, Baptiste Ristagno1,2, Julien Fonchastagner2, Nicolas Labbe1, Denis Netter2, Vincent Lanfranchi3, Noureddine Takorabet2

1VALEO, France; 2Laboratory GREEN, France; 3Université de Technologie de Compiègne, France

In the present work, an easily implementable method to reduce CPU time for some 3D FEM problems is presented. A translational

displacement between two parts on both sides of a planar airgap is taken into account with a change of coordinates in the airgap

region along the z-axis. To achieve this, formulation and post processing are slightly adapted. In this digest, the proposed method

is applied to an academic problem. Results are compared to a usual method. Time saving, ease of implementation and precision on

preliminary results are really encouraging with a strong reduction of CPU time keeping a low relative deviation. The method could

be easily applied to any configurations with planar airgap as MAGLEV, axial motor or levitation device.

Fast Linear Solver Based on Deflation and Proper Orthogonal Decomposition for Topology Optimization

Kota Watanabe, Kaito Oshima

Muroran Institute of Technology, Japan

A fast liner solver for topology optimization using a deflation technique with Proper Orthogonal Decomposition (POD) is discussed. The topology optimization method based on evolutionary algorithms such as genetic algorithm requires huge computational cost to evaluate many trial shapes. In this reason, a deflated Preconditioned Conjugate Gradient (PCG) method is introduced so as to reduce the cost of finite element analysis that is used to evaluate the objective function. The deflation technique decomposes the solution into fast and slowly converging components. The slow components can be solved by direct methods with low computational cost due to small dimensions. Therefore, the deflated PCC method can improve the convergence of PCG. However, the deflated PCG requires to find the slow components. In this study, a POD method with snapshots is introduced. In the evolutionary optimization process, solution vectors corresponding to parents are used for the snapshots. Orthogonal vectors for the deflation are constructed from the snapshots. In contrast with well-known model order reduction with POD, the present method has same accuracy of solution as the PCG method. Numerical results show that the present method can significantly reduce the computational cost of topology optimizations.

New Preisach Type Hysteresis Model Associated to XFEM in a HTS Cable modeling

Nana Duan1, Weijie Xu2, Shuhong Wang1, Jianguo Zhu3

1Xi'an Jiaotong University, China, People's Republic of; 2State Grid Shaanxi Electric Power Company Construction Branch, China, People's Republic of; 3The University of Sydney, Australia

In this paper, the new Preisach type hysteresis model is associated to improved extended finite element method (XFEM) in a high temperature superconducting (HTS) cable modeling. A coupled field-circuit analysis method for the HTS cable considering magnetic hysteresis by using the improved XFEM is presented. This hysteresis model is firstly combined with the improved XFEM to determine the magnetic hysteresis inductance. A magnetic field-circuit coupled program for current analysis of superconducting layers is coded. The numerical simulation results are reported compared with the experimental test results for the case of an HTS cable.

A Parallel Finite-Element Time-Domain Method for Nonlinear Dispersive Media

David S. Abraham, Dennis D. Giannacopoulos

McGill University, Canada

In this paper, a novel use of Graphics Processing Units (GPUs) is presented for the acceleration of Finite-Element Time-Domain (FETD) methods containing electrically complex media. By leveraging the massively parallel architecture of the GPU via NVIDIA’s Compute Unified Device Architecture (CUDA) language, the traditionally immense computational burden imposed by these materials can be largely alleviated, facilitating their modeling and incorporation into electromagnetic devices and systems. To that end, an analysis of the nonlinear dispersive FETD algorithm is presented in order to both identify computational bottlenecks and determine their amenability to parallelization. The performance of the resulting algorithm is then measured as a function of the simulation parameters, demonstrating up to a 200 times performance increase as compared to a traditionally serial implementation.

Convolutional Neural Network U-net applied in Transformer Multi-physics Analysis

Ruohan Gong, Zuqi Tang

Univ. Lille, Arts et Metiers Paris Tech, Centrale Lille, HEI, EA 2697 - L2EP - Laboratoire d'Electrotechnique et d'Electronique de Puissance, F-59000 Lille, France

With the development of power network, transformer is developing towards the trend of large capacity and extra-high voltage, then the ensuing stray losses caused by leakage magnetic flied is getting larger and larger, which not only lower the transmission efficiency of transformer but also leads to other problems such as local overheating, deformation and so on. The numerical simulation becomes an indispensable tool to enhance the design and optimization of transformer. However, the multi-physics computational cost of transformer can be very high due to its complex configuration. This paper puts forward a novel approach to analysis multi-physics problem for transformer with different winding structures based on deep learning model. With ground truth data obtained from finite element method (FEM) analysis, U-net, a convolutional neural network (CNN) is adopted to extract features and trained in supervised manner to predict the multi-physics analysis results for different topologies. During the training phase, inputs are FEM results which are represented in RGB and the outputs are evaluated by mean absolute error (MAE). The prediction results agree well with the ground truth. It is shown that the proposed approach can be seen as an effective tool to analysis multi-physics problem for transformer.