Modeling of broadband acoustic energy evolution in dynamic ocean environments using high‐performance computing resources.

Abstract

In recent years there has been considerable interest in predicting the acoustic energy temporal evolution over large 3‐D volumes. Most of the standard Navy models are inherently 2‐D and produce the acoustic pressure distribution of a point harmonic source in the vertical source‐receiver plane. There are several computational challenges that have to be addressed when using these models for quantitative modeling of the acoustic pressure distribution from natural or man‐made sources: broadband and complex spatial structure of a source, dynamic propagation environment, a need for 3‐D dynamic acoustic energy reconstruction, and an efficient visualization of a large amount of generated data. An attempt to resolve any of the mentioned issues pushes the computational time on a personal state‐of‐the‐art computer to several days (or weeks) and strongly dictates a need for upgrading these models to the high‐performance computing (HPC) environment. The range‐dependent acoustic model (RAM) was transferred into the HPC environment and upgraded to model broadband acoustic energy distribution from spatially distributed sources. The MPI domain decomposition approach to run the model on several cluster nodes is discussed. The data demonstrating computational time efficacy for multi‐node multi‐processor runs versus standard serial implementation are presented. [Research supported by TeraGrid Pathways Fellowship.]