论文部分内容阅读
在一定的规则剖分下，矩量法中的系数矩阵为Ｔｏｅｐｌｉｔｚ矩阵或多重Ｔｏｅｐｌｉｔｚ矩阵。利用这一特征而提出的共轭梯度（ＣＧＭ）和快速付里叶变换（ＦＦＴ）算法已成为目前国际上分析电大尺寸问题的一种有效手段。虽然ＣＧＭ—ＦＦＴ将普通矩量法中正比于Ｎ２（Ｎ为系数矩阵的阶数）的存储量压缩为正比于Ｎ的存储量，但其迭代算法使所花ＣＰＵ时间仍与普通矩量法相当。本文采用基于递推的Ｌｉｖｅｎｓｏｎ算法和一种库软件处理同样的问题，所花ＣＰＵ时间和普通矩量法相比降低两个量级，而且存储量比ＣＧＭ—ＦＦＴ技术还要小。本文以直导线的辐射和散射问题为例介绍了几种算法的基本原理，并对他们的计算时间和存储空间等进行了比较研究，得出了一些重要结论。
Under certain rules, the coefficient matrix in the method of moments is the Toeplitz matrix or the multiple Toeplitz matrix. Conjugate Gradient (CGM) and Fast Fourier Transform (FFT) algorithms proposed by using this feature have become an effective means to analyze the large size problem in the current world. Although CGM-FFT computes a memory volume proportional to N in normal moments method which is proportional to N2 (N is the order of the coefficient matrix), its iterative algorithm still consumes the same CPU time as the normal moment method . This article uses recursion-based Livenson algorithm and a library software to deal with the same problem, the CPU time spent compared with the general method of the moment reduced by two orders of magnitude, and storage capacity than CGM-FFT technology even smaller. In this paper, the basic principles of several algorithms are introduced by taking the radiation and scattering of straight lines as an example, and their computational time and storage space are compared and some important conclusions are drawn.