Classical And Modern Methods Of Deep Ocean Acoustic Tomography With Mobile Nodes

Acoustic wave is the main carrier of underwater communication.Before the study of depth measurement,underwater target positioning and other practical problems,it is necessary to obtain the underwater acoustic environmental parameters,among which the sound speed is one of the most important parameters.The sound speed is influenced by light,ocean currents and mesoscale vortices,which varies with time and space.By transmitting acoustic signals and analyzing the ocean environmental information carried by the received signals,ocean acoustic tomography can estimate the sound speed,which lays a foundation for the realization of long-term and large-scale ocean environmental observation.Mobile acoustic tomography reduces the experimental cost and improves the spatial resolution by increasing the spatial sampling,but correspondingly at the expense of temporal resolution.In this paper,the simulation study of three-dimensional mobile acoustic tomography is carried out by using the ocean data simulated by Hybrid Coordinate Ocean Model(HYCOM),and the data of the tomography experiment in the South China Sea in September 2020 is processed.The acoustic tomography problem is a typical problem of the underdetermined inverse problem,and the problem must be parameterized first.In this paper,Empirical Orthogonal functions(EOF)and Learned Dictionaries(LD)are introduced in the vertical dimension,and mesh division is introduced in the horizontal dimension to parameterize the three-dimensional sound speed field.There are often many solutions to an underdetermined problem.In order to select the "best"solution from many solutions,prior knowledge or constraint conditions need to be introduced.In this paper,two axiomatic methods of least square and maximum entropy are used to solve the inverse problem.They respectively add the minimum error square and the maximum entropy of solution as constraint conditions.In addition,this paper also tried the method of compressive sensing,in which the sound speed field was represented sparsely by LD,and the number of non-zero parameters was minimized as the constraint condition.In the simulation of three-dimensional mobile acoustic tomography,an area of 160km×160km in the South China Sea was taken as the target sea area.Mobile stations were placed around the target sea area and anchored sound sources are placed in the middle of the target sea area.Three inverse problem solving methods are applied to the three-dimensional moving acoustic tomog-raphy model to analyze and compare the errors and their advantages and disadvantages.All the three methods can estimate the overall trend of the sound speed field in the target sea area.The least square method is dependent on the prior knowledge and sensitive to the disturbance of time delay,but the tomography accuracy is higher after being added the prior knowledge.The maximum entropy method does not depend on prior knowledge,but the tomography error is large.The tomography error of the compressive sensing method is between the former two,and it does not depend on prior knowledge and is not sensitive to delay disturbance,but it will introduce observation matrix with large dimension.The influence of time-varying sound speed field caused by mobile acoustic tomography on sound speed tomography is also investigated.Mobile acoustic tomography can get the average of the sound speed field over a period of time,which will reduce the root mean square error,but reduce the temporal resolution at the same time.In the end,the actual data of the South China Sea acoustic tomography experiment in September 2020 are processed.Considering the distance dependent sound speed field,the three inverse problem solving methods used in the simulation are applied to the actual data processing.The overall root mean square error of the tomography results is all within lm/s,which verifies the feasibility of the acoustic tomography method for processing the experimental data.