利用拖船自噪声进行浅海环境参数贝叶斯反演

研究了以拖船自噪声为参考声源的浅海环境参数反演问题,并针对反演结果不确定性快速量化评估问题,提出了一种基于自适应重要性抽样的贝叶斯反演新方法。反演利用了拖船自噪声低频线谱成分,并采用混合高斯推荐函数自适应推荐声场模型样本,使得样本集中于参数高概率密度区域,实现后验概率密度快速收敛计算。仿真试验结果表明:拖船自噪声反演能够准确估计水深、沉积层及阵列参数等。所提自适应重要性抽样贝叶斯反演方法的计算效率优于快速吉布斯抽样方法。利用试验数据处理验证,反演得到试验海域声学环境参数,计算传播损失与各阵元接收线谱强度变化吻合,说明反演最优环境模型能准确表征声场传播特征。 

Bayesian geoacoustic inversion of self-noise in shallow water

The geoacoustic inversion problem based on tow-ship noise(self-noise) in shallow water is studied,and a Bayesian inversion method based on adaptive importance sampling is proposed to evaluate the inversion uncertainty.The tonal components at low frequencies of self-noise are utilized in the inversion.And the mixed gaussian function is used to propose the acoustic field samples adaptively to cover the high-probability distributed area of the parameters.The Posterior Probability Densities(PPDs) of parameters are estimated iteratively and effectively from the statistical analysis of acoustic field samples.The results of simulations show that,the water depth,seabed parameters and array parameters can be estimated accurately in self-noise inversion.Besides,the proposed method provides effective evaluations of PPDs while requiring significantly less acoustic field samples than Fast Gibbs Sampler(FGS).The method is verified through the processing of sea-trial data in a shallow water environment,and the geoacoustic parameters are inverted.The result shows that,the predicted Transmission Loss(TL) is consistent with the variations of spectral line intensity along the array,demonstrating the reliability of the inversion results.