Research on Bayesian Seabed Acoustic Parameter Inversion Method Based on Parallel Tempering Algorithm

摘要

Aiming at the problem of nonlinear inversion of seabed acoustic parameters under the semi-infinite elastic seafloor, the Bayesian inversion method is adopted, and the complex sound pressure at various distance points received by the hydrophone is used as the research object for inversion. Among them, the inversion parameters (including sound velocity, density, and sound velocity attenuation) are regarded as random variables, and the inversion results are given in the form of the posterior probability distribution, which can analyze the uncertainty of the inversion results. In order to speed up the inversion convergence, a parallel tempering algorithm is introduced, Markov chains at different temperatures are run at the same time so that the sampling process can be exchanged between Markov chains at different temperatures. This algorithm can effectively solve the common problems in Bayesian parameter inversion-a slow convergence caused by Markov Chain Monte Carlo (MCMC) algorithm. Numerical simulation experiments are used to compare the inversion effects of the simulated annealing (SA) algorithm and the parallel tempering algorithm. The simulation results show that compared with the simulated annealing algorithm, the parallel tempering algorithm can improve the inversion convergence speed and obtain the parameter inversion result with the smallest mean square error and the highest accuracy.