Abstract
In the context of the rapid development of computer hardware and the continuous improvement of the artificial intelligence and deep learning theory, aiming at the traditional numerical solution method to solve the underwater acoustic fluctuation equation with large computational volume and the limitation of using various acoustic propagation models. We use the numerical solution calculated by the KRAKEN based on the normal mode theory, which is widely used in low-frequency shallow water waveguides, and combine it with the idea of solving the retarded envelope function in the parabolic equation theory. We propose a physical information neural network (PINN)-based method for intelligent prediction of the acoustic field using the elliptic fluctuation equation as the controlling equation. We conduct experiments under water body sound velocity varying stratified waveguide, to validate the model forecasting effect. It is experimentally verified that an effectively trained PINN network model can forecast the sound field at any given range. The predicted sound field can be used for a wide range of applications, such as sound source localisation and sonar range estimation.