Range-dependent shallow water sound source localization via digital twin model incorporating spectral element full-wave numerical simulation

Sound source localization in complex range-dependent shallow water environments is a challenging problem. The challenges arise from factors such as irregular seabed topography and sound speed profiles, which cannot be quantified by the classical underwater sound propagation models. Although full-wave numerical simulation can accurately simulate the acoustic field propagation in such complex range-dependent environments, applying it directly to the sound source localization problem is difficult because of its high computational cost. To address this problem, this paper proposes an acoustic digital twin model that integrates full-wave numerical simulations in an efficient way to perform sound source localization both accurately and rapidly. Specifically, the spectral element method-based full-wave numerical simulations are used to model the sound propagation in a range-dependent environment for a small number of well-designed representations of sound source locations. The Kriging method is then employed to build a digital twin model based on these numerical results to form a model of the sound field varying continuously with source parameters. Subsequently, as long as the sound field is measured by a hydrophone array, it can be matched with the output acoustic field of the digital twin model based on the matched-field processing principle to realize rapid and accurate source localization in complex underwater environments. Numerical experiments demonstrate that the proposed method offers accurate and low-cost real-time sound source localization in complex marine scenarios.