Machine learning methods for estimating probability density functions of transmission loss: Robustness to source frequency and depth

Abstract

Predicted values of transmission loss (TL) in ocean environments are sensitive to environmental uncertainties. The resulting predicted-TL uncertainty can be quantified via the probability density function (PDF) of TL. Monte Carlo methods can determine the PDF of TL but typically require thousands of field calculations, making them inappropriate for real-time applications. Thus, a variety of alternative techniques based on polynomial chaos, field shifting, modal propagation in ocean waveguides, and spatial variations of TL near the point(s) of interest have been proposed. Recently, an approach to estimating the PDF of TL based on nominal TL, ocean environmental parameters, and machine learning was found to have a success rate of 95% with constant source depth (91 m) and frequency (100 Hz) when tested on 657,775 receiver locations within 100 randomly selected ocean environments. This presentation describes an extension of this approach and its success predicting the PDF of TL for different source depths and frequencies for ranges up to 100 km. This increase in the size of the parameter space furthers the need for a sophisticated method of choosing training examples. Such a method is proposed, and its performance is compared to that of prior techniques. [Work supported by ONR.]