A large-scale parallel method and numerical simulation for 3D explosion problems

The numerical simulation of three-dimensional (3D) explosion problems is typically characterized by a large computational scale, making it difficult to complete the calculations using serial computation on a single computer. Currently, parallel computing based on clusters is more suitable for dealing with such problems. In this paper, a distributed parallel architecture involving a dual shared memory parallelism strategy is developed to address issues related to high communication overhead and low parallel efficiency in large-scale computing of 3D explosion problems. In this strategy, a non-uniform memory access (NUMA) architecture is built, a shared memory window is created using the distributed message passing interface (MPI), and a complete computing system is constructed on the Linux platform for the 3D Eulerian multi-material parallel hydrocode PMMIC-3D [1]. Then, the experiments of the free-field near-earth air-explosion are carried out, the simulations using the above parallel computational strategy are performed in this work, and the peak overpressure results of simulations are in agreement with experimental data. Furthermore, the computing time and parallel efficiency are analyzed by the numerical simulations of 3D multi-point air-explosion in a wide-area field. The analysis results indicate a substantial reduction in computing time and an improvement in parallel efficiency when compared to previous methods at the same computational scale. Moreover, by varying the reference positions and masses in the simulations, the attenuation law and overpressure-distance curves of the explosion shock wave are derived, offering valuable data to support both the assessment of explosion-related incidents and the optimization of rescue operations within complex urban environments.