Evaluating applicability of an azimuthally symmetric model in a laboratory tank with absorptive panels

Abstract

In our lab at Brigham Young University, a tank of water will be used to help validate machine learning algorithms that could be applied to ocean acoustics. A model for sound propagation in the tank is required, and we are exploring the possibility of using ORCA, a range-independent normal mode model [Westwood et al., JASA (1996)]. Because ORCA assumes azimuthal symmetry in cylindrical coordinates, reflections from the tank walls must be significantly attenuated for ORCA to be a good model. To compare our measurements to ORCA, two approaches have been tried: (1) timegating and (2) absorptive panels on the walls. The layering used in ORCA is air on top, water with a maximum depth of 0.91 m, 3 cm thick acrylic bottom, a lower air layer, and then concrete on bottom. A simpler model excluding the air between the acrylic and the concrete was also tested. Both models were compared to measured data; both relative and absolute transmission loss were compared across a large frequency band and as a function of source/receiver distances at set depths. If ORCA is a good model for the tank, then we could easily generate training data for machine learning applications.