A Parallel Computing Strategy for Large-scale Numerical Simulation of Explosion Field Based on the Eulerian Method

doi:10.12382/bgxb.2025.0338

Abstract

Abstract:

To address the issues of large computational scale and low parallel efficiency in the numerical simulation of explosion fields, a distributed shared memory parallel computing strategy based on the Eulerian method is proposed. This strategy improves the parallelization of the pMMIC3D program by constructing a cluster system with a non-uniform memory access (NUMA) architecture and using the message passing interface (MPI) and InfiniBand (IB) high-speed network, thereby enhancing the large-scale computational capability of program. The accuracy, speedup ratio, parallel efficiency and computational scale of the parallel program are tested through the simulation of the explosion in the air. The results show that the proposed parallel computing strategy is accurate and effective, significantly reducing the communication overhead and improving the computational efficiency. In addition, the applicability of this parallel computing strategy in complex scenarios is further validated through the explosion test inside a concrete building structure, and the calculated results are compared with experimental data. The calculated results show that this parallel computing strategy has the ability to handle the numerical simulation of complex large-scale explosion field and is suitable for practical engineering applications.

Key words: explosion field, Eulerian method, parallel strategy, non-uniform memory access, message passing interface

NING Jianguo, GAO Yi. A Parallel Computing Strategy for Large-scale Numerical Simulation of Explosion Field Based on the Eulerian Method[J]. Acta Armamentarii, 2025, 46(10): 250338-.

Figures/Tables 18

References 28

[1]	刘迪, 陈菁, 张安强, 等. 爆炸冲击波作用下聚脲材料对肺冲击伤防护作用的数值模拟研究[J]. 爆炸与冲击, 2024, 44(12):84-94.
	LIU D, CHEN J, CHEN A Q, et al. Numerical simulation study on the protective effect of polyurea materials against lung impact injuries under the action of blast shock waves[J]. Explosion and Shock Waves, 2024, 44(12):84-94. (in Chinese)
[2]	唐皓, 刘云龙, 冯集团, 等. 水下爆炸异相气泡动力学特性的Euler有限元数值模拟研究[J]. 应用数学和力学, 2023, 44(8):895-908.
	TANG H, LIU Y L, FENG J T, et al. Numerical simulation of the dynamics of heterogeneous bubbles in underwater explosion by Euler finite element simulation[J]. Applied Mathematics and Mechanics, 2023, 44(8):895-908. (in Chinese)
[3]	张子寒, 赵跃堂, 胡康. 基于LBE方法对Euler空气网格进行特定冲击波荷载的设定[J]. 防护工程, 2021, 43(5):28-34.
	ZHANG Z H, ZHAO Y T, HU K. Setting of specific shock wave loads for Euler air grids based on the LBE methodology[J]. Protective Engineering, 2021, 43(5):28-34. (in Chinese)
[4]	XIONG M, XU C, GAO X, et al. Improved grid partitioning algorithms for load-balancing high-order structured aerodynamics simulations[J]. Computers and Electrical Engineering, 2018, 67:70-84. doi: 10.1016/j.compeleceng.2018.03.016 URL
[5]	JIANG Z, TAKAYAMA K, MOOSAD K, et al. Numerical and experimental study of a micro-blast wave generated by pulsed-laser beam focusing[J]. Shock Waves, 1998, 8(6):337-349. doi: 10.1007/s001930050126 URL
[6]	FOX G C, WILLIAMS R D, MESSINA P C. Parallel computing works![M]. Morgan Kaufmann Publishers, 1994.
[7]	W E BAKER, P S WESTINE, et al. Explosion hazards and evaluation[M]. Netherlands: Elsevier Science, 1983.
[8]	BAO X, LI S, LIU J, et al. 3D multiscale analysis method for explosion problems based on the substructure of the explosion source[J]. Engineering Structures, 2022, 252:113633. doi: 10.1016/j.engstruct.2021.113633 URL
[9]	蒋新生, 陈日, 余彬彬, 等. 耦合化学动力学机理的油气爆炸燃烧特性分析[J]. 油气储运, 2022, 41(6):12.
	JIANG X S, CHEN R, YU B B, et al. Characterisation of oil and gas explosion and combustion by coupled chemical kinetic mechanism[J]. Oil and Gas Storage and Transport, 2022, 41(6):12. (in Chinese)
[10]	王鑫, 张祖雨. 基于神威加速计算架构的LBM多级并行计算[J]. 计算机系统应用, 2024, 33(8):60-67.
	WANG X, ZHANG Z Y, LBM multi-Stage parallel computing based on Shenwei accelerated computing architecture[J]. Computer Systems and Applications, 2024, 33(8):60-67. (in Chinese)
[11]	QI X F, GUO L X, TSANG H. MPI-based parallel FDTD for EM scattering from coated complex targets[J]. Sessionap, 2010, 1(1):432-435.
[12]	赵金庆, 马天宝. 基于伪弧长移动网格算法的爆炸与冲击多介质问题数值模拟[J]. 兵工学报, 2020, 41(S2):200-210.
	ZHAO J Q, MA T B. Numerical simulation of explosion and shock multi-media problems based on a pseudo arc-length moving grid algorithm[J]. Acta Armamentarii, 2020, 41(S2):200-210. (in Chinese)
[13]	宁建国, 李钊, 马天宝, 等. 动能弹高速侵彻钢筋混凝土靶时弹丸头部质量侵蚀微观机理[J]. 兵工学报, 2021, 42(9):1809-1818. doi: 10.3969/j.issn.1000-1093.2021.09.001
	NING J G, LI Z, MA T B, et al, Microscopic mechanism of mass erosion of projectile head during high speed penetration of kinetic projectile into reinforced concrete target[J]. Acta Armamentarii, 2021, 42(9):1809-1818. (in Chinese)
[14]	CHENG W, YONG B, WEN H, et al. Large-scale parallel computing for 3D gaseous detonation[C]// Proceedings of the International Conference on Parallel Computing in Fluid Dynamics. Berlin,Germany: Springer, 2014:541-552.
[15]	XU J, SMITH T J, FERNANDEZ M, et al. Massive data storage and sharing algorithm in distributed heterogeneous environment[J]. Journal of Intelligentand Fuzzy Systems, 2018, 35(4):4017-4026.
[16]	ATMAJA E H, SONAWANE K. Distributed computing and shared memory-based utility list buffer miner with parallel frameworks for high utility itemset mining[J]. InternationalJournal of Business Intelligence and Data Mining, 2023, 23(2):125-149.
[17]	MILTON J, ZARKESH HA P. Impacts oftopology and bandwidth on distributed shared memory systems[J]. Computers, 2023, 12(4):86. doi: 10.3390/computers12040086 URL
[18]	牟洪锋. 基于共享存储的分布式数据库查询下推策略研究[D]. 武汉: 华中科技大学, 2020.
	MOU H F. Research on push down strategy for distributed database query based on shared storage[D]. Wuhan: Huazhong University of Science and Technology, 2020. (in Chinese)
[19]	PANIGRAHI G, KODALI N, PANDA D, et al. Fast hardware-aware matrix-free algorithms for higher-order finite-element discretized matrix multivector products on distributed systems[J]. Journal of Parallel and Distributed Computing, 2024, 192:104925. doi: 10.1016/j.jpdc.2024.104925 URL
[20]	TOMAZL B P, JULIA M S, FARIA M P. ADABA: improving the balancing between runtime and accuracy in a new distributed version of the alpha-beta algorithm[J]. Artificial Intelligence Review, 2023, 56(5):4255-4293. doi: 10.1007/s10462-022-10269-3
[21]	KUMAR K P. Distributedshared memory: a page-based approach[J]. international journal of scientific research in engineering and management, 2023, 7(6):21454.
[22]	NING J G, MA T B, FEI G L. Multi-material Eulerian method and parallel computation for 3D explosion and impact problems[J]. International Journal of Computational Methods, 2014, 11(5):1350079. doi: 10.1142/S0219876213500795 URL
[23]	DASHTI M, FEDOROVA A, FUNSTON J, et al. Trafficmanagement: A holistic approach to memory placement on NUMA systems[J]. ACM SIGPLAN Notices, 2013, 41(1):381-394.
[24]	COMPTON D A, CLEVELAND D W. NUMA, a nuclear protein involved in mitosis and nuclear reformation[J]. Current Opinion in Cell Biology, 1994, 6(3):343-346. doi: 10.1016/0955-0674(94)90024-8 pmid: 7917323
[25]	LAMETER C. Anoverview:NUMA becomes more common because memory controllers get close to execution units on microprocessors[J]. Queue, 2013, 11(7):2508834.
[26]	SZUSTAK L, WYRZYKOWSKI R, KUCZYNSKI L, et al. Architecturaladaptation and performance-energy optimization for CFD application on AMD EPYC Rome[J]. IEEE Transactions on Parallel and Distributed Systems: A Publication of the IEEE Computer Society, 2021, 32(12):2852-2866.
[27]	MA T B, REN H L, LI J, et al. Large scale high precision computation for explosion and impact problems[J]. Chinese Journal of Theoretical and Applied Mechanics, 2016, 48(3):599-608. doi: 10.6052/0459-1879-15-382
[28]	李军润, 卢永刚, 冯晓伟, 等. RC箱型结构内爆炸载荷特性和动力行为分析[J/OL]. 爆炸与冲击, 2025[2025-01-14]. https://www.bzycj.cn/article/doi/10.11883/bzycj-2024-0388.
	LI J R, LU Y G, FENG X W, et al. Analysis of explosion load characteristics and dynamic behaviour within RC box structures[J/OL]. Explosion and Shock Waves, 2025[2025-01-14]. https://www.bzycj.cn/article/doi/10.11883/bzycj-2024-0388. in Chinese)

算例	计算域/cm	网格步长/cm	网格数量
测试1	200×200×1000	1.0	200×200×1000
测试2	200×200×1000	0.78	256×256×1282
测试3	200×200×1000	0.62	323×323×1613

算例	计算域/cm	网格步长/cm	网格数量
测试1	200×200×1000	1.0	200×200×1000
测试2	200×200×1000	0.78	256×256×1282
测试3	200×200×1000	0.62	323×323×1613

网格步长	分区方式	子区域网格数	P₁峰值/ GPa
	1×1×1	100×100×100	0.0812
	3×1×1	34×100×100	0.0810
0.2×0.2×0.2	1×3×1	100×34×100	0.0808
	1×1×4	100×100×25	0.0801
	3×4×1	34×25×100	0.0799
	4×4×4	25×25×25	0.0796
	1×1×1	400×400×400	-
	3×1×1	134×400×400	0.0811
0.05×0.05×0.05	1×3×1	400×134×400	0.0799
	1×1×4	400×400×100	0.0795
	3×4×1	134×100×400	0.0803
	4×4×4	100×100×100	0.0806

网格步长	分区方式	子区域网格数	P₁峰值/ GPa
	1×1×1	100×100×100	0.0812
	3×1×1	34×100×100	0.0810
0.2×0.2×0.2	1×3×1	100×34×100	0.0808
	1×1×4	100×100×25	0.0801
	3×4×1	34×25×100	0.0799
	4×4×4	25×25×25	0.0796
	1×1×1	400×400×400	-
	3×1×1	134×400×400	0.0811
0.05×0.05×0.05	1×3×1	400×134×400	0.0799
	1×1×4	400×400×100	0.0795
	3×4×1	134×100×400	0.0803
	4×4×4	100×100×100	0.0806

关键点位置/m	公式峰值/MPa	仿真峰值/MPa	误差/%
1.2	0.188	0.165	12.2
1.6	0.102	0.090	11.8
2.2	0.055	0.049	10.9
2.4	0.047	0.042	10.6
2.6	0.041	0.037	9.8