隐重启动精化Arnoldi缩减金属导体部分单元等效电路模型  

Implicitly Restarted-Refined Arnoldi Method to Reduce PEEC Model of a Conductor

作　　者：张玉廷 蔡智[2] 张华[1] 刘胜[3] 

机构地区：[1]中国空间技术研究院总体部,北京100094 [2]工业和信息化部软件与集成电路促进中心,北京100038 [3]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《电工技术学报》2013年第12期83-89,共7页Transactions of China Electrotechnical Society

基　　金：总装预研基金资助项目(9140A33040212HT0501)

摘　　要：针对金属导体部分单元等效电路模型渐进波形估计缩减方法的数值病态和普通Arnoldi缩减算法受初值选择影响和收敛性复杂的问题,提出了基于隐重启动精化Arnoldi算法的金属导体PEEC模型缩减方法。将主导极点的抽取转化为主导特征值的抽取。对系统主导特征值精化,并通过重启动抑制非主导特征值。在此基础上,对所建立的金属导体模型进行简化,给出了简化后不同支路的电阻和电感值。通过分析和算例对比证明了算法的有效性。通过对一毫米级导体阻抗频率特性进行仿真和试验对比验证了算法的正确性。仿真阻抗幅频特性最大误差为4.1%,相频特性最大误差为2.7%。算法可有效克服传统模型缩减算法只能提取有限缩减支路和缩减支路较少时提取极点质量不高的不足,提高了工程应用价值。The implicitly restart-refined Arnoldi method is proposed to reduce the partial element equivalent circuit model of a conductor, which is to solve the numerical ill-conditioning of asymptotic waveform evaluation model reduction method and the complex astringency of common Arnoldi method influenced by initial value. The new algorithm transfers the extracting of dominant poles to pick up corresponding dominant eigenvalues. The dominant eigenvalues are refined, while the non-dominant eigenvalues are restrained by restarting the algorithm. On the basis of above, the resistance and inductance of different reduced decoupled branches are obtained. The algorithm is effective by analyzing and an example comparison. And the correctness is also proved by simulating and testing the impedance-frequency characters of a millimeter-level conductor. The max simulation amplitude- frequency character error is 4.1%, while the max phase-frequency character error is 2.7%. The algorithm avoids the defects of only drawing several reduced branches and picking up low quality poles as fewer branches of the traditional methods. It is a most significant in en^ineerinm

关 键 词：部分单元等效电路模型 隐重启动精化Arnoldi算法 金属导体 主导特征值 频 率响应 

分 类 号：TM93[电气工程—电力电子与电力传动]