Research On Shallow Sea Layered Acoustic Parameter Inversion Method Based On Backward Feedback Neural Network Model

The parameters describing the acoustic characteristics of the seafloor are defined as"Geoacoustic parameters",including medium sound velocity(including acoustic attenuation)and density.The acquisition of acoustic parameters in shallow Haiti has always been a classic and hot topic in the field of underwater acoustics.However,the above mentioned acoustic parameters are difficult to be measured directly and in a large range,so the accurate inversion of the above parameters by using acoustic technology has long attracted the attention of researchers in related fields,which has important research significance and application value.In this thesis,In order to get close to the actual shallow sea environment,the shallow sea bottom is regarded as a multi-layered layered structure,and the seabed model with elastic layered structure is established.A method of acoustic parameter inversion based on the multi-layered elastic structure based on the backward feedback neural network model is studied.Firstly,the theoretical prediction value of the shallow sea sound pressure Field was obtained by Fast Field Method(FFM).Secondly,the relationship model between the predicted sound pressure field and the ground sound parameter values to be retrieved was established according to the BPNN model,and the neural network inversion model under different hierarchical structures was trained.Then,the ground sound parameters under this layer were estimated by combining the sound field model.Finally,the measured sound pressure field data is introduced into the neural network model to obtain the inversion results.By comparing with simulated annealing algorithm,the uncertainty of inversion results is analyzed.The inversion method in this thesis is studied through simulation analysis and laboratory pool shrinkage test,and the research conclusions are as follows:1.Compared with the existing acoustic parameter inversion studies in shallow Haiti,which are carried out on the premise of seabed stratification determination,it is impossible to accurately invert sound pressure information under unknown seabed conditions.The inversion method of earth acoustic parameters based on BPNN model studied in this thesis makes the whole neural network model quickly approximate the mapping relationship between measured data,hierarchical structure and earth acoustic parameters to be inversion by adjusting the weight and threshold of neurons in the constructed model,which can effectively judge the seabed stratification and meet the same accuracy requirements as the traditional optimization algorithm.And the determined neural network model can be directly used to solve the same type of problems,avoid repeated calculation,greatly improve its application efficiency and application prospects.2.The analysis of the influence of acoustic parameters on acoustic propagation characteristics shows that:for the acoustic pressure field in water,the five acoustic parameters of shallow sea bottom density,p-wave sound velocity,S-wave sound velocity,p-wave sound velocity attenuation and S-wave sound velocity attenuation have different influences.The uncertainty of sound velocity and density inversion is smaller than sound velocity attenuation,and the influence of layer thickness on sound propagation characteristics is similar to that of sound velocity.The accuracy of the inversion results of S-wave acoustic velocity c_p,p-wave acoustic velocity c_s and the densityρ_b of the sedimentary layer are all higher than the two attenuation coefficients:S-wave attenuationα_p and p-wave attenuationα_s,which accords with the physical property that the first three types of acoustic parameters have a greater influence on the sound field information of shallow sea in the forward modeling.3.In the inversion calculation,the classical simulated annealing algorithm is used respectively,compared with the neural network model established in this thesis in terms of computational efficiency and inversion results.Under the premise of setting the same accuracy,the number of iterations of BPNN model is reduced to 2%of SA algorithm,and the computational efficiency of inversion results can be increased by 70%.The error between inversion results and preset truth values under simulation conditions is less than 3%,and the applicability of BPNN model in acoustic parameter inversion in shallow Haiti is well verified by pool test data.4.The shallow sea bottom model is constructed under the experimental conditions of equal ratio reduction.The neural network model completed by training is respectively used for single-layer and double-layer experimental model inversion calculation,which realizes the judgment of submarine sound field information on the seabed layered structure,and the earth sound parameter inversion under the corresponding layered model.The experimental results have a high degree of agreement with the prior parameter results.Based on the algorithm system of self-learning and constant correction of neural network model,it has a wide application prospect in the shallow geoacoustic parameter inversion algorithm of the same type.