The Construction of Sound Speed Field Based on Back Propagation Neural Network in the Global Ocean

Abstract

The sound speed is a key parameter that affects the underwater acoustic positioning and navigation. Aiming at the high-precision construction of sound speed field in the complex marine environment, this paper proposes a sound speed field model based on back propagation neural network (BPNN) by considering the correlation of learning samples. The method firstly uses measured ocean parameters to construct the temperature and salinity field. Then the spatial position, the temperature and the salinity information are used to construct the global ocean sound speed field based on the back propagation neural network algorithm. During the processing, the learning samples of back propagation neural network are selected based on the correlation between sound speed and distance. The proposed algorithm is validated by the global Argo data as well as compared with the spatial interpolation and the empirical orthogonal function (EOF) algorithm. The results demonstrate that the average root mean squares of the BPNN considering the correlation of learning samples is 0.352 m/s compared to the 1.527 m/s of EOF construction and the 2.661 m/s of spatial interpolation, with an improvement of 76.9% and 86.8%. Therefore, the proposed algorithm can improve the construction accuracy of sound speed field in the complex marine environment.