Underwater Sound Speed Field Forecasting Based on the Least Square Support Vector Machine

Junting Wang,Tianhe Xu,Wei Huang,Liping Zhang,Jianxu Shu,Yangfan Liu,Linyang Li

DOI: https://doi.org/10.3390/jmse12030480

IF: 2.744

2024-03-14

Journal of Marine Science and Engineering

Abstract:Underwater sound speed is one of the most significant factors that affects high-accuracy underwater acoustic positioning and navigation. Due to its complex temporal variation, the forecasting of the underwater sound speed field (SSF) becomes a challenging task. Taking advantage of machine learning methods, we propose a new method for SSF forecasting based on the least square support vector machine (LSSVM) and a multi-parameter model, aiming to enhance the forecasting accuracy of underwater SSF with hourly resolution. We first use a matching extension method to standardize profile data and train the LSSVM with the parameters of observation time, temperature, salinity, and depth. We then employ radial basis function kernels to construct the forecasting model of SSF. We validate the feasibility and effectiveness of the LSSVM model by comparing it with the polynomial fitting (PF) and back propagation neural network (BPNN) methods, using hourly data obtained from the measured data and open data. The results show that the means of the root mean square for the LSSVM based on the observation time parameter and the LSSVM based on the multi-parameter model achieve 0.51 m/s and 0.45 m/s, respectively, presenting a significant improvement compared with the PF (0.82 m/s) and BPNN (0.76 m/s) methods.

oceanography,engineering, marine, ocean

What problem does this paper attempt to address?

The paper primarily focuses on addressing the prediction problem of the Sound Speed Field (SSF) underwater, particularly improving prediction accuracy at an hourly resolution. Specifically, the paper proposes a new method based on the Least Square Support Vector Machine (LSSVM) to predict SSF and utilizes various parameter models to enhance prediction accuracy.

To tackle this problem, the authors first standardized the profile data using the matching extension method and trained the LSSVM by observing parameters such as time, temperature, salinity, and depth. Then, a radial basis function kernel was used to construct the SSF prediction model. To verify the effectiveness and feasibility of the proposed LSSVM model, the authors compared it with Polynomial Fitting (PF) and Back Propagation Neural Network (BPNN) methods. The results show that the LSSVM based on the observation time parameter and the LSSVM based on the multi-parameter model achieved Root Mean Square Errors (RMSE) of 0.51 m/s and 0.45 m/s, respectively, significantly better than the PF method (0.82 m/s) and the BPNN method (0.76 m/s). This indicates that the proposed LSSVM method can more accurately predict SSF, contributing to the improvement of underwater positioning, navigation, timing, and communication systems.