The Sound Speed Profile Inversion And Its Software Realization Base On RBF Neural Network

The ocean Sound Speed Profile(SSP)determine the propagation of underwater acoustic signals.Because the SSP varies with the dynamic change of the sea water,Timely and accurate acquisition of the real-time SSP of the sea water has great impacts on the detection and monitoring of marine,the evaluation of sea water,the exploitation of submarine resources,the positioning of submarines and the deployment of seabed forces.In a word,the ocean SSP is one of the key points in the study of underwater acoustics.In this paper,the inversion method of SSP in the sea area is studied,and the limitation of the inversion method based on historical data is analyzed.This paper presents radial basis function neural network(RBFNN)based on and the existing research,which is applied in inversion of SSP.The RBFNN is used to map complex nonlinear relationship between the surface temperature,historical data,the stability feature of the sea area and the actual SSP,due to its nonlinear processing ability.But in the parameters of the hidden layer center,width and output weight trained separately and its shortcoming of local minima.The genetic algorithm(GA)is used to optimize the three parameters of the RBF neural network.The coding method put all the parameters in one chromosome optimize them synchronously.It can strengthen the cooperation between the hidden layer and the output layer,and obtained the global optimal parameters.Aiming at the defects of the genetic algorithm,the genetic algorithm is improved to ensure the global optimization ability of the genetic algorithm.In this paper,the model is simulated and verified by the Argo observation data from 2006 to 2017 in a region of the South China Sea.The variation of SSP in this area is analyzed and studied.The characteristics of the stability law of the velocity profile of the sea water are obtained,and the measured data show that the SSP in the sea area fluctuates in this stable feature.On this foundation,the GA-RBF inversion method proposed and applied to simulate the model combining with the measured temperature and historical data in the sea area.With the mean square root error as the evaluation index,the data of June and December of 2016 to 2017 in the region are tested.By comparing with other methods,the average root mean square error of two years is obtained by this method is 26.93%of AVG method,50%of EOF method,63.30%of BP method,and 66.34%of no optimization for RBF method.The simulation results demonstrate that based on the GA-RBF network model and using the inversion results from the measured surface temperature of the sea area,the measured velocity profile is closer to the measured velocity profile,and the model can be used for the inversion of the vertical SSP of the sea area.Using the fitting ability of RBF neural network,learning ability and approximation ability of nonlinear function,the accuracy and scope of inversion of SSP can be improved.Finally,based on the Lab VIEW software as the platform and the SQL Server 2012 as database,and using the modular design idea,a visual velocity profile inversion software is designed and developed.The software has friendly interface and convenient operation.It is easy for users to study and operate,and is easy to upgrade and maintain.The development of the software can be used as the basic research tool for underwater acoustic signal processing.The retrieved data obtained can be used as the input parameter of the underwater acoustic signal processing algorithm model,and it has very important engineering application value.