PINNs for Sound Propagation and Sound Speed Field Estimation Simultaneously

Ocean dynamics phenomena can induce structural changes in sound speed, which, in turn, affect sound propagation in both depth and range directions. Typical methods for estimating underwater wavefields rely on traditional models or numerical solvers, which require complete sound speed information that is difficult to obtain. Recently, Physics-informed Neural Networks (PINNs) have emerged as a promising approach for solving physical problems, they have been applied to various problems but nothing about 2D underwater sound propagation. In this study, we propose a PINNs model composed of two independent networks to simultaneously estimate the sound propagation and sound speed field. Through simulations, we have demonstrated that PINNs can yield excellent results and tackle problems that traditional methods cannot solve. The estimated sound speed field can be utilized to model disturbances and gain insight into spatial changes in the ocean, which has practical application value and exciting prospects.