Study On Seafloor Sediment Parameter Inversion Based On Acoustic Backscattering Model

In the development and application of ocean,the investigation of seabed sediment can meet the needs of marine engineering and research.Meanwhile,seabed are also an important boundary of ocean waveguide,which has an important impact on the performance of sonar system.In order to further improve the theoretical basis of sediment survey and the performance of geoacostic inversion method,this paper introduced a more efficient and accurate posterior probability density(PPD)sampling method in the Bayesian theory,inverting the backscattering strength of sediment to geoacoustic parameters and input parameters of geoacoustic models with quantified uncertainty.The specific work is as follows:(1)Seabed high-frequency scattering is generally divided into roughness scattering and volume scattering,and the difference of their contributions to total scattering is closely related to the sediment.Therefore,the relative entropy of prior to posterior in bayesian inversion is proposed to estimate the dominated backscattering mechanism.Diffe Rential Evolution the Adaptive Metropolis is fist introduced to geoacoustic inversion to sample the parameter posterior space more efficiently and accurately,and the relative entropy of prior to posterior in bayesian inversion is proposed to estimate the dominated backscattering mechanism:roughness scattering from the sediment interface,volume scattering from the sediment interior,and mixed scattering from both.Simulation results show that the proposed method can correctly estimate the dominated scattering mechanism.(2)Different media assumptions(fluid,elastic or poroelastic media)for sediments corresponds to different geoacoustic models and acoustic scattering models.However,a single media assumptions or model cannot be appropriate for all types of sediments,and it is difficult to select the best model for inversion especially in the absence of prior information.In this paper,the combination of multiple models and model selection method are first introduced to geoacoustic inversion to improve the applicability of inversion method and obtain the best model and inversion results for the measurements.The bayesian factors between different models is given through multiple quantified model selection criteria,and the results show that:for the current meassurements of backscattering strength of sandy sediment,the geoacoustic model and scattering model based on the poroelastic medium is considered to be the best model.(3)In order to reduce the computational cost of multi-model combination based inversion,the substitute model is adopted to further improve the efficiency of geoacoustic inversion.By referring to the idea of iterative construction of substitute model in posterior space of parameters,this paper applies multiple-try DREAM_(ZS)(MT-DREAM_(ZS))to geoacoustic inversion for the first time,and an adaptive Gaussian process model-based Bayesian geoacoustic inversion(AGMB-GI)based on MT-DREAM_(ZS)was proposed.At the same time,adaptive resampling(AR)strategy was introduced and AGMB-GI-AR was proposed to improve the updating efficiency of the substitute model.The experimental results showed that both methods can get similar inversion results compared to the original model,but the bayes factor comparison shows that AGMB-GI-AR can obtain similar results with AGMB-GI faster,and the final results are very close to the inversion results based on the original model.(4)Based on the method proposed in this paper,two sets of backscattering strength which are measured at Quinault site and Chudao are used for inversion,and the results show that the PPDs obtained by inversion are consistent with the limited measurements.The results of model selection show that the best model for the two sets of data are both the combination of Effective Density Fluid Model(EDFM)and fluid acoustic scattering model.Compared with the inversion results of the data at Quinault site,the difference between the three models decreases in the inversion results of the data at Chudao.The results of relative entropy estimation show that the dominated scattering mechanism of the sandy sediments at Quinault site is roughness scattering,while the energy of roughness scattering and volume scattering of the clay sediments at Chudao are similar.In addition,inversion of backscattering strength data at more frequencies and grazing angles can further reduce the uncertainty of the inversion results.In fact,the inversion results given in the form of PPD under the framework of bayesian theory are essentially the most reasonable and complete solution to the inverse problem based on the priori information,limited data and the selected model.