Determining A Geoacoustic Model From Shallow Water Refraction Profiling And Transmission Loss Data, Using A Simulated Annealing Inversion Algorithm

摘要

To characterise the propagation conditions in a shallow-water environment at low frequencies, measurements have been made of both CW transmission loss (TL) versus distance at 4 frequencies from 50 to 250 Hz, and travel times of airgun-generated head waves. The head wave data yield the sound-speed and time-intercept of a reflecting interface, and these results are used as known parameters when the CW data are inverted to obtain a complete geoacoustic model. The inversion algorithm was an adaptive implementation of the simulated annealing method [W L Goffe et al, Journal of Econometrics, 60. 65 - 99 (1994)]. For this particular environment it was found that the geoacoustic model should consist of 3 uniform solid layers overlying a solid uniform basement. The cost function is the RMS of the residuals between measured and OASES-modelled values for TL over the 4 frequencies; and the objective is to determine optimum values for the 5 elastic parameters and the thickness of each layer. To reduce the number of unknown parameters, regression equations were devised to relate the 2 less critical parameters (density and shear absorption) to the 3 that are usually found to be more important (sound and shear speeds, and sound absorption). With 11 unknown parameters (the head wave data yield one thickness), the inversion algorithm found a satisfactory geoacoustic model and, after nearly 34000 runs, gave indications that the result is likely to be the global optimum.