摘要
针对计算电磁学算法应用于实际工程问题时面临的计算模型电大尺寸多尺度瓶颈,基于自研的高性能并行计算框架JASMIN,通过提出融合h-自适应时间积分算法流程的对称半正定亚网格FDTD算法流程,实现了具备大规模并行能力的亚网格时域有限差分(FDTD)算法.该算法保留了FDTD算法高并行可扩展性优势,同时可局部使用细化网格剖分电大尺寸模型精细结构部位,支持粗网格区域和细网格区域分别采用大时间步长和小时间步长进行迭代计算.算例表明,该方法在解决电大多尺度电磁模拟问题时,具有较强的并行可扩展性,同时能够保证数值计算结果的精度和稳定性,与传统的FDTD算法相比,可显著降低内存需求,提高计算效率.
In this paper,due to the presentation of an algorithm procedure combined with h-adaptive integration procedure based on the high-performance parallel computing framework JASMIN,a symmetric positive semi-definite subgridding finite difference time domain(FDTD)method was developed with the ability of large-scale parallel computation,overcoming the bottleneck problems of computational efficiency in engineering practice,and making FDTD method achieve large-scale parallel computation capability.Retaining the advantage of good parallel expansibility of FDTD,the arithmetic was arranged to dissect the fine structure of a large-scale model based on refined locally meshes,and support the iterative computation simultaneity with smaller time and longer time steps separately in the region of different refined meshes.Numerical examples show that the proposed method has good parallel scalability,and can ensure the accuracy and stability of numerical results.Compared with the traditional FDTD method,it can significantly reduce the memory requirements and improve computational efficiency.
作者
陈晓洁
鲍献丰
李瀚宇
张爱清
周海京
CHEN Xiaojie;BAO Xianfeng;LI Hanyu;ZHANG Aiqing;ZHOU Haijing(Software Center for High Performance Numerical Simulation,China Academy of Engineering Physics,Beijing 100088,China;Institute of Applied Physics and Computational Mathematics,Beijing 100094,China)
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2023年第6期649-656,共8页
Transactions of Beijing Institute of Technology
基金
国家自然科学基金资助项目(12001053,62231003)。
关键词
电大多尺度模型
对称半正定亚网格FDTD算法
大规模并行
electrically large and multi-scale model
positive semi-definite subgridding FDTD method
largescale parallel computation
