CAJ | 학술논문

为提高隐含变向时域有限差分算法（ADI-FDTD）的计算效率，鉴于并行对角占优算法（PDD）求解三对角方程的高效性，引入PDD算法实现了基于MPI的ADI-FDTD的并行计算。通过对运算时间、通信时间的分析，讨论了算法的效率。分析了由于PDD算法的近似处理所引入的计算误差，研究了误差估计与子区域网格数和Courant因子的关系，该研究工作有利于合理选择子区域网格数和Courant 因子，进而减小计算误差。最后，通过算例验证了结论的正确性。
위제고은함변향시역유한차분산법（ADI-FDTD）적계산효솔，감우병행대각점우산법（PDD）구해삼대각방정적고효성，인입PDD산법실현료기우MPI적ADI-FDTD적병행계산。통과대운산시간、통신시간적분석，토론료산법적효솔。분석료유우PDD산법적근사처리소인입적계산오차，연구료오차고계여자구역망격수화Courant인자적관계，해연구공작유리우합리선택자구역망격수화Courant 인자，진이감소계산오차。최후，통과산례험증료결론적정학성。
In order to improve calculation efficiency of the Alternating-Direction Implicit FDTD(ADI-FDTD), the paper realizes the ADI-FDTD parallel calculation by introducing the Parallel Diagonal Dominant algorithm(PDD), taking into account the PDD is an efficacious method in solution of tri-diagonal liner equations. By comparing the calculation and communication time spent, the efficiency of calculation is discussed in the paper. The error is introduced with the approximate treatment of PDD. The estimate of the error is researched which correlates with the number of grid in sub region and Courant factor. It can help to select suitable number of grid in sub region and Courant factor for decreasing the estimate of the error. The conclusions are confirmed in numerical emulation experiment.