A. Caiti firstname.lastname@example.org
DIST - Universitá di Génova, Via Opera Pia 13, IT-16145, Génova, Itália
S.M. Jesus email@example.com
Universidade do Algarve, Campus de Gambelas, PT-8000, Faro, Portugal.
Statoil, Trondheim, Norway
Comments: download pdf.
Ref.: IEEE Journal of Oceanic Engineering, Vol. 21, No. 4, pp. 355-366, 1996.
AbstractAcoustic propagation in shallow water is greatly dependent on the geo acoustic properties of the sea-bottom. This paper exploits this dependence for estimating geoacoustic sediment properties from the bottom acoustic returns of known signals received on a hydrophone line array. There are two major issues in this approach: one is the feasibility of acoustic inversion with a limited aperture line array, the other is related to the knowledge of the geometry of the experimental configuration. To test the feasibility of this approach, a 40 hydrophone - 4 m spaced towed array together with a low-frequency acoustic source, was operated at a shallow water site in the Strait of Sicily. In order to estimate the array deformation in real time, it has been equipped with a set of non-acoustic positioning sensors (compasses, tiltmeters, pressure gauges). The acoustic data was inverted using two complementary approaches: a genetic algorithm (GA) like approach and a radial basis functions (RBF) inversion scheme. More traditional methods, based on core sampling, seismic survey and geophone data, together with Hamilton's regression curves, have also been employed on the same tracks, in order to provide a ground truth reference environment. The results of the experiment, can be summarized as follows: 1) the towed array movement is not negligible for the application considered, and the use of positioning sensors are essential for a proper acoustic inversion, 2) the inversion with GA and RBF are in good qualitative agreement with the ground truth model, and 3) the GA scheme tends to have better stability properties. On the other hand, repeated inversion of successive field measurements requires much less computational effort with RBF.