Three-dimensional off-grid localization of incipient tip vortex cavitation using Bayesian inference

Abstract

In this paper, we propose a 3-D source localization method to determine the incipient tip vortex cavitation (TVC) positions by constructing a signal model for the broadband off-grid cavitation. Subsequently, the conventional sparse Bayesian learning (SBL), whose performance is diminished owing to an inappropriate signal representation for the off-grid TVC by basis mismatch, was extended to solve the signal model via Taylor series and the block-sparse based broadband off-grid approach (off-grid BSBL). The off-grid BSBL was examined using synthetic and experimental cavitation tunnel data. Furthermore, its localization results were compared with those for conventional and adaptive beamformers and on-grid BSBL, which generally suffer from low resolutions and basis mismatch, respectively. When the off-grid BSBL was applied to chirp- and pop-type signals from incipient TVC, it determined the TVC positions near the top center of the propeller and provided the clearest localization results among the schemes considered.