Detection of an internal solitary wave by the underwater vehicle based on machine learning

Abstract

A new hydrodynamic artificial intelligence detection method is proposed to realize the accurate detection of internal solitary waves (ISWs) by the underwater vehicle. Two deep convolution neural network structures are established to predict the relative position between the underwater vehicle and ISW and the flow field around the underwater vehicle. By combining field observation data and the computational fluid dynamics method, accurate numerical simulation of the motion of the underwater vehicle in a real ISW environment is achieved. The training process for the neural network is implemented by building a dataset from the above results. It is shown that the position prediction accuracy of the network for ISW is larger than 95%. For the prediction of the flow field around the underwater vehicle, it is found that the addition of the convolutional block attention module can increase the prediction accuracy. Moreover, the reduction of the number of sensors by the dynamic mode decomposition method and k-means clustering method is realized. The accuracy can still reach 92% even when the number of sensors is reduced. This study is the first to use hydrodynamic signals for the detection of ISW, which can enhance the navigation safety of underwater vehicles.