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
Session
PC-M1: Numerical Techniques
Time:
Thursday, 18/Jul/2019:
10:50am - 12:40pm

Session Chair: João Pedro Assumpção Bastos
Session Chair: Frédéric Guyomarch
Location: Patio 44-55

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

PEEC modeling of planar spiral resonators

Dimitri Voltolina1, Riccardo Torchio2, Paolo Bettini2,3, Roberto Cavazzana3, Maurizio Moresco3

1Centro Ricerche Fusione, University of Padova, Italy; 2Department of Industrial Engineering, University of Padova, Italy; 3Consorzio RFX, Italy

Planar spiral resonators match the main requirements of compactness and efficiency for their application in RF plasma sources as excitation coils. The electrodynamic of these resonators is not trivial and requires appropriate numerical models in view of plasma source design. In this paper we present a Partial Element Equivalent Circuit model of a sample spiral resonator, which is solved to explore its resonant modes.



Fast Methods for Speeding up Induction Machine Simulation

Chuan Lu, Dingsheng Lin, Ningning Chen, Bo He, Ping Zhou

Ansys, United States of America

In induction machine simulation, it usually takes quite a long time to reach the steady state due to the large time constant. In this paper, two methods are proposed to speed up the transient process to reach the steady state. In both methods, the initial condition of the simulation is estimated from the solution of the locked FEA model with equivalent conductivity/resistance. The effectiveness of

these methods is validated by two examples.



Space-Time Finite-Element Eddy Currents Analysis for a Surface-Mounted Permanent Magnet Machine

Allaa Eddine Boumesbah, Guillaume Krebs, Claude Marchand, Frédéric Bouillault

GeePs | Group of electrical engineering - Paris

This paper presents an analysis of eddy currents distribution through the cross section of windings of a surface-mounted permanent magnet machine (PMSM). A time-discretized field-circuit coupled finite-element (FE) model associated with a mortar technique based model reduction is used to perform accurate and fast losses calculation. Copper losses behavior is analyzed for a no-load test, where the influence of machine speed and magnet position is observed.



LOD-Homogenization of High Contrast Eddy Current Problem

Xiaotao Ren1,2, Antti Hannukainen3, Anouar Belahcen1

1Department of Electrical Engineering and Automation, Aalto University, Espoo, Finland; 2Integrated Actuators Laboratory, Ecole Polytechnique Fédérale de Lausanne, Neuchâtel, Switzerland; 3Department of Mathematics and Systems Analysis, Aalto University, Espoo, Finland

We have built a localization orthogonal decomposition homogenization for multiscale eddy current problem. The subspace projector depends on the property contrast of the materials. To deal with this problem, in this study, we propose an approach of a divergence-free

gauging condition instead of an artificial conductivity regularization.



Effect of winding distribution in stator slots on eddy current losses

Loucif Benmamas1, Guillaume Krebs1, Rachid Missoum2, Ahmed Lakab3, Claude Marchand1, Eduardo Santander3, Damien Maroteaux2

1GeePs | Group of electrical engineering - Paris; 2Renault Group; 3Altran

The work presented in the present paper deals with computation of eddy current distribution in machine windings. The calculation method is based on Finite Element model coupled with time-discretized circuit. A model reduction technic based on mortar method is used to perform fast calculation. The method is applied on wound rotor synchronous machine in a full load configuration. Analysis of eddy currents shows that current density distribution among windings cross sections, and so losses depend on turns positioning in stator slots



Hybrid Method for Solving 2D Steady-State Scattering Problems of Electromagnetic Wave: Application of Collocation EFGM-BEM

Ayumu Saitoh, Teruou Takayama, Atsushi Kamitani

Yamagata University, Japan

The hybrid method of the collocation element-free Galerkin method and the boundary element method is applied to the 2D steady- state scattering problems of electromagnetic wave. In addition, the performance of the proposed method is investigated numerically. The results of computations show that the relatively smooth distribution of electric field is obtained for both internal and external regions regardless of the wave number. Therefore, it is found that the proposed method can be used as one of tools for solving the 2D steady-state scattering problems of the electromagnetic wave.



A Novel Electrical Impedance Tomography Post-processing Scheme based on the Denoising Autoencoder Neural Network

Nasir Hussain, Charles T. M. Choi

National Chiao Tung University, Taiwan

Electrical Impedance Tomography (EIT) is a non-invasive technique for imaging. It allows us to generate a conductivity distribution profile based on the measured voltage from the test subject boundary when an electrical current is injected into the test subject. Various types of EIT inverse solvers have been used to reconstruct images, but their results are unsatisfactory for diagnosis purposes because of low image resolution. Neural network-based methods have been used to enhance EIT images. In this paper, we present a new EIT post-processing scheme based on the denoising auto-encoder neural network (dNANN) to reconstruct EIT images with better results. This neural network, an EIT post-processor, is applied to a 2D finite element model following an EIT inverse solver to train the neural network in the presence of noise. The results show significant improvement in EIT image quality in the presence of noise.



Divide and Conquer: A Novel Approach for Multi-targets in Electrical Impedance Tomography

Nasir Hussain, Charles T. M. Choi

National Chiao Tung University, Taiwan

Electrical Impedance Tomography (EIT) has been used as an imaging tool for many clinical and industrial applications. EIT images are created from voltages that are collected from the boundary of a body when an electrical current is injected into it. The change in conductivity is calculated from the differential voltage measured at the boundary. One major EIT reconstruction issue is multiple-target problems: EIT imaging technique performance degrades significantly for multiple targets inside a body. To overcome this problem, we propose a divide-and-conquer method in conjunction with neural networks as part of an EIT post-processing scheme. After an EIT image is created, it is divided into multiple parts and post-processed using a neural network technique. Afterward, the sub-images are merged to form a whole image. The results show that the proposed scheme delivers a significant improvement over images reconstructed using EIT inverse solvers for multiple-target problems.



Residual Type a posteriori Error Estimates for 3-D Low Frequency Stable Maxwell Formulations in Both Frequency and Time Domains

Zuqi Tang1, Yanpu Zhao2

1Univ. Lille, Arts et Metiers Paris Tech, Centrale Lille, HEI, EA 2697 - L2EP - Laboratoire d'Electrotechnique et d'Electronique de Puissance, F-59000 Lille, France; 2Wuhan University, School of Electrical Engineering and Automation, Wuhan, China

In this paper, residual type a posteriori error estimates developed in our previous work for magnetostatic and eddy current problems are extended to low-frequency (LF) Maxwell problems in $\mathbf{A}/\varphi$ formulation. Both inductive and capacitive effects can be handled in the studied frequency- and time- domain formulations. High order finite element basis (HOFEB) and low order finite element basis (LOFEB) of edge and nodal elements are adopted in numerical examples to evaluate the performance of the proposed estimator.



Tree gauging in lossy high frequency FEM models

Gergely Koczka2, Oszkar Biro1

1IGTE, TU-Graz, Austria; 2Transformers Weiz, Siemens Inc. Austria

Wave propagation problems involving both high conductivity materials and large non-conducting domains are solved in the frequency domain by the method of finite elements (FEM) using edge basis functions and applying the A,V-formulation. The singular matrix of the resulting algebraic equation system is regularized by tree gauging to facilitate its direct solution. It is shown that choosing a random tree can result in erroneous solutions even if a highly sophisticated sparse parallel direct equation system solver is used. This problem is overcome by generating the tree by a special algorithm taking account of the presence of high conductivity materials. Two numerical examples are investigated: an academic 1D wave propagation problem and a real-word 3D antenna.



A Broadband Enhanced Nodal-Order Reduction Methodology for Large-Scale Equation Sets of Three-Dimensional Transient Field Problems

Jiajia Chen1, Shiyou Yang1, Zhi Gong1, Zhuoxiang Ren2

1College of Electrical Engineering, Zhejiang University, China; 2Sorbonne Universités, UPMC Univ. Paris 06, France

In the numerical analysis of three-dimensional (3D) transient field problems, the order of equation sets is extremely higher. Moreover, the long-time span of a transient process further deteriorates this situation in case of transient behavior computations using a numerical approach. Consequently, different model order reduction (MOR) techniques are proposed and applied in different areas of physics in order to reduce the computational cost of a large-scale system. However, most exiting MOR techniques cannot be used effectively for broadband frequency transient analysis. In this regard, a broadband Enhanced Nodal-Order Reduction (ENOR) methodology is proposed and validated using the numerical results as reported.



A Network Topological Approach based Transient Three-Dimensional Electro-Thermal Model of Insulated Gate Bipolar Transistor

Jiajia Chen1, Shiyou Yang1, Zhi Gong1, Zhuoxiang Ren2

1College of Electrical Engineering, Zhejiang University, China; 2Sorbonne Universités, UPMC Univ. Paris 06, France

With the continuously miniaturizing and rapidly increasing in power ratings of an insulated-gate bipolar transistor (IGBT), the exact junction temperature becomes one of the critical performance parameters in evaluating the reliability of the transistor. In this regard, the network topological method is extended to develop a physics-based transient electro-thermal model, considering both the three-dimensional (3D) heat complications and the nonlinearity of the thermal parameters, of an IGBT. The network topology method uses the graph theory to describe the network according to the network's geometric structure. Due to its clear physical meaning, easiness in numerical implementation, and extremely high efficiency in numerical computation, the proposed 3D model has an extraordinarily high computational efficiency as compared with the finite element method, and more accurate as compared with the one-dimensional lumped RC model, as evidenced by the numerical results as reported.



Structure-Preserved POD for Parametric Low Frequency Fullwave Problems with Gauged Potential Formulations

Shuai Yan1,2, Zuqi Tang3, Thomas Henneron3, Zhuoxiang Ren1,4

1Institute of Electrical Engineering, Chinese Academy of Science, 100190, Beijing, China; 2Institute of Microelectronics, Chinese Academy of Science, 100029, Beijing, China; 3Univ. Lille, Arts et Metiers Paris Tech, Centrale Lille, HEI, EA 2697 - L2EP - Laboratoire d'Electrotechnique et d'Electronique de Puissance, F-59000 Lille, France; 4Sorbonne University, UR2, L2E, F-75005 Paris, France

The proper orthogonal decomposition (POD) is applied for multi-dimensional parametric analysis in low frequency problems considering both capacitive and inductive effects. We use a recently proposed symmetric gauged formulation involving both magnetic vector potential and electric scalar potential. A structured-preserving strategy for POD is implemented since the calculation involves different physical quantities associated with different geometric entities. A numerical example is provided to illustrate the capability of POD in deriving multi-dimensional parametric models.



Time Domain Finite Element Method for Near Field Computations of Low Frequency Metamaterials

Zhi Gong1, Shiyou Yang1, Qing Huo Liu2

1Zhejiang University, China, People's Republic of; 2Duke University, USA

To simulate the dynamic performance of devices including low frequency magnetic metamaterials, a time domain finite element method (TDFEM) is derived. The magnetic properties of metamaterial is modelled as a bulk one, and approximated by using a standard Lorentz model, and 2D case studies are solved by using different methods. The accuracy of the proposed method is validated by comparing its results with those of the frequency domain method. Unlike most existing works, algorithms and software which mainly focus on the metamaterials excited by electromagnetic wave, this work studies metamaterials in magnetoquasistatic field and facilitates future analysis of magnetic metamaterial applications in power engineering.



Linear Interpolation-based Method with Branch Connection Constraints for Movement in Reluctance Network Model

Hongqin Xie1, Guillaume Krebs2, Maya Hage-hassan2, Man Zhang2, Claude Marchand2, Zhuoxiang Ren3

1Sichuan University, China, People's Republic of; 2Group of electrical engineering Paris (GeePs),France; 3UPMC, Laboratoire d’Électronique et Électromagnétisme (L2E), France

This paper proposes a new method to obtain smooth movement in reluctance network model (RNM). In which, linear interpolation-based method (LIM) is introduced to guarantee the continuity of the node magnetic potentials on the sliding interface. Besides, additional branch connection constraints (BCC) are put forward to ensure the continuity of branch magnetic fluxes along the sliding interface. With both LIM and BCC, the computation accuracy of the movement in RNM is improved. The validation of the proposed LIM-BCC is verified in comparison with linear interpolation method (LIM) and fixed-step method (FSM).



Modeling of scalar dependencies of soft magnetic material magnetization for electrical machine finite element simulation

Pengfei Zhang1,2, Xiao Xiao1, Fabian Müller1, Gregor Bavendiek1, Nora Leuning1, Silas Elfgen1, Jun Zou2, Kay Hameyer1

1Institute of Electrical Machines (IEM), RWTH Aachen University, Germany; 2State Key Laboratory of Power System, Department of Electrical Engineering, Tsinghua University, Beijing, PR China

The magnetization characteristic of silicon steel corresponds significantly diverse to its parameters, such as amplitude, angle between flux desnisty B and magnetic field strength H, magnetizing frequency of excitation and mechanical stress. Those factors occur locally in electrical machines and influence their performance. Therefore, it is crucial to measure the material characteristics under realistic conditions for adequate identification. In this paper, modeling and simulation of soft magnetic material are performed based on a detailed study of measured data considering magnetization amplitude, angle, frequency and stress. The dependent soft magnetic material characteristics are derived from the measurement data and concluded into interpolation surfaces. For instance, these surfaces are used in the IEM finite element software iMOOSE to simulate an example, here a synchronous machine with tooth-wound stator coils and surface rotor magnets. A challenge is the accurate determination of the local material properties depending on the point of operation, which influence the global quantities, such as losses and torque. This paper provides a methodology to consider various locale influences on silicon steel in a finite element simulation, offering the potential for improved electromagnetic circuit design.



Partial Least Square Regression for Electromagnetic Acoustic NDT Probe

Houssem Boughedda1, Tarik Hacib1, Mostafa Kamel Smail2, Hakim Sadou1, Yann Le Bihan3, Hulusi Acikgoz4

1Laboratoire d’électronique et d’électrotechnique industrielle (L2EI); 2Institut Polytechnique des Sciences Avancées; 3Laboratory GeePs; 4Engineering Faculty, Karatay University

The aim of this paper is to propose a new efficient and reliable approach on the field of Non Destructive Testing (NDT), for the diagnosis of detects in non-ferromagnetic material by Electromagnetic Acoustic Transducer (EMAT). EMAT is a recent ultrasonic technique that generates and detects ultrasonic waves in conductive material without physical contact. A forward model based on Finite Element Method (FEM) is developed to simulate the EMAT response for a given materials in order to build a database for the inversion tool. The Partial Least Square (PLS) Regression (PLSR) is used to solve the inverse problem in order to detect and characterize defects in non-ferromagnetic materials as a fast, simple and accurate inversion tool. It is a dimensionality reduction method which aims to model the relationship between the matrix of independent variables (predictors) X and the matrix of dependant variables (response) Y. The purpose of PLS is to find the Latent Variables (LV) having the higher ability of prediction by projecting original predictors into a new space of reduced dimension. A set of results is carried out in order to validate the calculations.



Parallelism Implementation and Branch Optimization of EBE-FEM Based on CUDA Platform

Yan Zhang, Xiuke Yan, Sheng Wang, Dongyang Wu, Baodong Bai, Bin Xia

School of Electrical Engineering, Shenyang University of Technology, Shenyang, 110870, China

The finite element analysis of large complex structures makes higher demand on memory capacity and computation speed, which leads to the inefficiency of traditional serial finite element method (FEM) for such large-scale problems. EBE-FEM is a parallel algorithm on element level, which avoids the formation of the total coefficient matrix and requires less memory capacity, and can achieve good parallelism. In this paper, The EBE-FEM has been implemented parallelly on CUDA platform, and been programmed using C++ language. The thread branches exist in protocol operation have been researched and optimized to improve parallel efficiency. The correctness of EBE-CG method is verified by the finite element analysis of an open slot of motor. The optimized program is applied to analyze the main magnetic field of single-phase transformer. The results show that the EBE-FEM implemented on CUDA platform is more effective than serial EBE-FEM, and branch optimization can improve the speedup ratio further.