A novel parallelized model for simulating massive underwater three-dimensional acoustic fields at medium-high frequencies

Underwater three-dimensional acoustic fields can provide more accurate results than two-dimensional one. However, three-dimensional simulations are computationally intensive and conventional models cannot meet the application demands. A parallelized model for massive underwater three-dimensional acoustic field simulation is proposed in this paper. First, by optimizing variable storage structure and introducing field loops, a serial model is developed based on three-dimensional ray theory, which is capable of simulating massive acoustic fields. Second, a hybrid parallel computing strategy for the serial model is designed at field loops and α loops based on a combination of open multiprocessing and message passing interface, which has both the communication advantage of open multiprocessing and the scalability advantage of message passing interface. Furthermore, a dynamic scheduling algorithm between processes is proposed, which significantly optimizes load balancing by allowing the control process to continuously monitor the computing processes and dynamically distribute tasks. Finally, two sets of numerical simulation experiments, each containing 243 three-dimensional acoustic field computation tasks, are conducted in the Pacific and Atlantic Oceans based on the global ocean reanalysis dataset from 2021. Two experiments demonstrate that the novel parallelized model achieves speed up of 271.76 and 278.65 with 384 cores, which are 6.20 and 5.85 times faster than conventional model, respectively.