In order to efficiently solve large-scale asymmetric dense linear matrix in discrete field integral equations, a preconditioned restarted changing minimal residual method based on the Hessenberg process (CMRH) is proposed. It is used to implement the modified multilevel fast multi-pole algorithm (MLFMA) to compute the scattering problem of perfect electric conductors in a lossy half space. MLFMA is used to accelerate matrix vector multiplication of the CMRH. Scattering characteristics of the cylinder, the box and the missile model are presented. Numerical results show that the CMRH method can efficiently reduce both iteration number and convergence time as compared to the generalized minimal residual (GMRES). Furthermore, CMRH is more easily combined with sparse approximate inverse (SAI) and symmetric successive over relaxation (SSOR) preconditioning techniques, making it more practical.%为了高效求解半窄间三维电磁散射问题中离散电场积分方程产生的大型对称稠密复线性矩阵,将半空间多层快速多极子方法与CMRH方法相结合,其中多层快速多极子方法用十加速CMRH方法中的矩阵矢量乘运算.为了验证文中方法的有效性,分别计算了位于有耗半空间的圆柱体、长方体以及某导弹模型的散射特性.结果表明,所提出出的方法不仅可以滤除高频误差,平滑低频误差,而且能使求解半空间离散电场积分方程的迭代次数和计算时间比现在广泛使用的广义最小余量法显著减少.同时,CMRH方法与稀疏近似逆预条件、对称超松弛预条件结合可进一步提高求解效率.
展开▼
