Research On Geoacoustic Inversion By Single Vector Hydrophone Based On Nonlinear Bayesian Theory

With more and more attention paied on the marine field in China,understanding and mastering the marine environment information becomes an urgent issue.Because acoustic properties such as the sound speed,density,and attenuation in the sea bottom have important influence on the sound propagation in the underwater channel,acquisitioning the sediment properties through acoustic inversion technology has become one of the important topics in underwater acoustic research.With the developments during low frequency and very low frequency(VLF)in underwater acoustics research,to analyze and verificate acoustic propagation problems often require deeper seafloor information.Therefore,it is more and more urgent to study acoustic inversion techniques for deeper seafloor properties.Therefore,this paper makes use the nonlinear Bayesian theory to invert deeper sediment parameters using the sound pressure and vertical particle velocity recorded by a VLF vector hydrophone and the method is tested using the sea trial data.Firstly,a global optimization algorithm—adaptive simplex simulated annealing Algorithm(ASSA)is studied,which combines the advantages of simulated annealing algorithm and downhill simplex algorithm.And it is suitable to use this optimization algorithm to invert the sediment parameters.The wave impedance characteristics of the acoustic vector field in shallow water are studied,which shows that the vertical wave impedance carries more sediment information and is very suitable to invert sediment properties,in comparing with the horizontal wave impedance.A parallel program is developed for rapid calculation of the sound field,and then a nine-parameter inversion scenario of a multi-layer seafloor scenario is conducted to test ASSA algorithm.The numerical simulation shows that ASSA algorithm has a good capability in globally searching and can satisfy the need in geoacoustic inversion.Secondly,the nonlinear Bayesian inversion theory is studied,after which the Markov chain Monte Carlo(MCMC)method is studied,including rejection sampling method,Metropolis algorithm,transitional Markov chain Monte Carlo method(TMCMC),adaptive annealing Metropolis algorithm(ATM).Numerical simulations are conducted to test the ability of ATM in generating random samples of high-dimensional probability density functions.Based on the same nine-parameter inversion scenarios of multi-seafloor scenarios,ATM was used for geoacoustic inversion and the maximum a posteriori(MAP)estimation of the seafloor parameters was obtained,and the correlation characteristics among the inversion parameters and uncertainty are analyzed.Finally,geoacoustic inversion based on the nonlinear Bayesian theory was carried out for the data recorded by a vector sensor in South China Sea.The actual seafloor parameters of some certain sea area were obtained,in which the mutual relationship among inverted parameters and the uncertainties is analyzed qualitatively and quantitatively.