An experiment on inversion of sound profile through ray modeling

Abstract

An experiment was conducted in a lake of 150-m depth. A very short hydrophone array (3-m long or 1/50 of the water depth) was used for data acqisition, which consists of 13 hydrophones with 25-cm spacing. Receiving signals are modeled as sums of eigenrays determined mainly by the sound profile, while the sound profile is expressed by six measured samples at different depths in addition to a known sound speed at the bottom. The inversion is based on a simulated annealing algorithm off-line using a 486/100 microcomputer and completed within 2 min. The resultant rms error of sound profile is about 1.6 m/s in comparison with the measured data. An important feature of this approach is that instead of matching the measured to the modeled data vector, arriving structure expressed by the DOA angles and relative time delays are matched for the inversion. To estimate the arriving structure, high-resolution spectrum analysis techniques such as OWSF and MUSIC are adopted. The experiment has shown that inversion of the sound profile can be successfully achieved through ray modeling and using a very short vertical array.