环境-声场不确定性传递过程中代理建模方法

Novel Surrogate Modeling Method in Uncertainty Propagation from Environment to Acoustic Field

海洋环境参数的不确定性是声场预报不确定性的主要来源之一。海洋环境参数和声场之间通常具有非常强的非线性关系,导致通过海洋环境参数的不确定性来计算声场的不确定性较为困难。传统使用的蒙特卡罗方法,需要多次运行声场计算模型,导致计算量过大。针对此,文中提出了一种新的基于多项式-克里金方法(PC-Kriging)的代理建模方法,可以高效地计算海洋环境参数至声场之间的不确定性传递过程。该方法使用多项式方法来提取系统响应的全局趋势,使用克里金方法逼近局地响应。同时,基于标准失配测试模型进行了计算机仿真,从模型逼近精度以及传播损失概率密度函数(PDF)2个方面进行验证。结果表明:PC-Kriging代理建模方法在逼近精度上优于单独使用多项式或者克里金方法,传播损失PDF也与蒙特卡罗方法结果吻合得较好,适合用于环境-声场不确定性传递过程计算。

Uncertainty of ocean environmental parameters is one of the main sources of uncertainty in underwater acoustic field prediction.In addition,the relationship between the environmental parameters and the acoustic field may be highly nonlinear,so it is difficult to calculate the uncertainty of acoustic field through the uncertainty of ocean environmental parameters.The traditional Monte Carlo method needs to run the acoustic field calculation model many times,which results in too much calculation.To overcome the drawbacks of the existing methods,this paper proposes a new surrogate modeling method based on polynomial-chaos-Kriging(PC-Kriging)which can calculate the uncertainty transfer process efficiently from ocean environmental parameters to acoustic field.This method extracts the global trend of the system response using the polynomial chaos expansions(PCE)and approaches the local response using the Kriging method.Computer simulations using the general mismatched benchmark acoustic environmental model is performed to verify the proposed method in two aspects-the accuracy of the model approaching accuracy and the probability density function of the propagation loss.The results indicate that the surrogate modeling method based on the PC-Kriging is efficient and better in accuracy than the PCE or Kriging method.The probability density functions of propagation loss obtained by PC-Kriging and direct Monte Carlo are consistent,showing the suitability of PC-Kriging for the uncertainty propagation from environment to acoustic field.