Fast Large-Scale Electromagnetic Field Analysis Using PC Cluster for Optimal Design of Electric Machines

使用 PC 集群进行快速大规模电磁场分析以实现电机优化设计

基本信息

批准号：
14550265
负责人：
SHIMASAKI Masaaki
金额：
$ 2.24万
依托单位：
Kyoto University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2002
资助国家：
日本
起止时间：
2002 至 2003
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-14550265/
关键词：
algebraic multigrid (AMG) method finite edge-element method parallel ICCG method block red-black ordering parallel genetic algorithm magnetic hysteresis stop model 代数マルチグリッド法連立一次方程式電磁界解析並列計算代数的マルチグリッド法 H行列遺伝的アルゴリズム最適化問題 /

项目摘要

Precise analysis of complex phenomena and detailed design optimization of electric and electronic devices require a large-scale electromagnetic field analysis, which costs a large amount of computation time. However, the recent development of high performance computing techniques including parallel computation has not yet been applied sufficiently to the large-scale electromagnetic field analysis.The purpose of this research is to achieve the fast and precise large-scale electromagnetic field analysis for optimal design of electric machines under parallel computing environment. For this purpose this research aims to develop (1) efficient algorithms for fast electromagnetic field analyses, and (2) advanced optimal design methods for electric machines including accurate iron-loss evaluation.The obtained results are categorized into the four elements below.(1-a)Algebraic Multigrid (AMG) Algorithm for Electromagnetic Field AnalysesAn AMG solver for linear systems of equations involving real symmetric H-matrices is developed. A new AMG technique for electromagnetic finite edge-element analysis is proposed.(1-b).Parallel Solvers for Linear Systems of EquationsA new parallel ordering, "block red-black ordering," is proposed for a parallelized ICCG solver with fewer synchronization points and a high convergence rate.(2-a)Parallel Genetic Algorithm (GA) for Optimal Design of Electric MachinesThe island parallel GA efficiently achieves an optimum design of electric machinery on a distributed parallel computing system.(2-b)Magnetic Property Modeling and Nonconforming Mesh FEM for Advanced Electromagnetic Field AnalysisEfficient arid accurate DC hysteresis models for electrical steel sheets are developed, which are based on the stop and/or play models. The DC hysteresis models are used for eddy-current analyses to represent AC B-H loops accurately. The mortar finite element method is studied, which is applied to a magnetic-field analysis of rotating machinery.

对复杂现象的精确分析和电气电子设备的详细设计优化需要大规模的电磁场分析，这需要大量的计算时间。然而，近年来包括并行计算在内的高性能计算技术的发展尚未充分应用于大规模电磁场分析。本研究的目的是实现快速、精确的大规模电磁场分析，以优化设计并行计算环境下的电机。为此，本研究旨在开发（1）用于快速电磁场分析的有效算法，以及（2）先进的电机优化设计方法，包括精确的铁损评估。所获得的结果分为以下四个要素。（1 -a) 用于电磁场分析的代数多重网格 (AMG) 算法开发了一种用于涉及实数对称 H 矩阵的线性方程组的 AMG 求解器。提出了一种用于电磁有限边缘元分析的新 AMG 技术。(1-b).线性方程组的并行求解器提出了一种新的并行排序“块红黑排序”，用于同步点较少的并行 ICCG 求解器(2-a)电机优化设计的并行遗传算法岛式并行遗传算法在分布式并行计算上有效地实现了电机的优化设计(2-b)用于高级电磁场分析的磁特性建模和非相容网格有限元开发了基于停止和/或运行模型的高效且精确的电工钢板直流磁滞模型。直流磁滞模型用于涡流分析，以准确表示交流 B-H 环。研究了砂浆有限元法在旋转机械磁场分析中的应用。