柱体FDTD算法中的特殊处理方法及其应用研究

    Research on Special Processing Techniques and Applications of FDTD Algorithm for Cylindrical Structures

    • 摘要: 对比旋转对称柱体,时域有限差分(FDTD)算法在非对称柱体下对应由不同扇体累加的周期循环结构,需进行电/磁场量和接入电/磁流在角向的差分离散处理。同时,数学上的奇异性使轴线的电/磁场量计算亦需重点关注。为降低数值色散误差,首先,从FDTD算法的基本时域方程出发,利用周期性循环思想,分别对FDTD电/磁场方程和惠更斯体电/磁流在角向的差分离散进行处理。其次,以FDTD电/磁场差分离散求解过程是否溢出为判断标准,若该过程中某电/磁场方程分母中存在径向网格点为零的情况,则需对该电/磁场方程进行轴上单独处理。进而利用斯托克斯定理推导出待轴上处理的电/磁场量方程。基于散射体的Mur边界、匹配层、电磁波带隙和负折射完美透镜算例证明,角向和轴上的处理可有效降低非对称柱体下FDTD算法的数值色散误差,且对计算效率无大的影响。

       

      Abstract: Compared with the body of revolution cylinder, the finite-difference time-domain (FDTD) algorithm for asymmetric cylinders corresponds to a periodic cyclic structure accumulated by different fan-shaped segments, which requires the azimuthal difference discretization processing of electric/magnetic field quantities and connected electric/magnetic currents. Meanwhile, mathematical singularities make it necessary to focus on the calculation of electric/magnetic field quantities on the axis. To reduce numerical dispersion errors, firstly, starting from the basic time-domain equations of the FDTD algorithm, the periodic cycle idea is used to process the azimuthal difference discretization of FDTD electric/magnetic field equations and Huygens volume electric/magnetic currents respectively. Secondly, taking whether overflow occurs in the FDTD electric/magnetic field difference discretization solution process as the judgment criterion, if the denominator of an electric/magnetic field equation contains a zero radial grid point during this process, separate on-axis processing is required for this electric/magnetic field equation. Then, Stokes′ theorem is used to derive the electric/magnetic field equations to be processed on-axis. Examples including scatterers with Mur boundaries, matched layers, electromagnetic band gaps and negative-refraction perfect lenses prove that the azimuthal and on-axis processing can effectively reduce the numerical dispersion errors of the FDTD algorithm for asymmetric cylinders and have no obvious influence on computational efficiency.

       

    /

    返回文章
    返回