A Novel Acoustic Method for Cavitation Identification of Propeller

When a propeller is under state of cavitation, it will experience negative effects, including strong noise, vibration, and even damage to the blades. Accordingly, detection cavitation has attracted attention researchers. Propeller noise signal contains wealth information, which powerful method identify state. Considering nonlinear characteristics feature describing complexity signals, called refined composite multiscale fluctuation-based dispersion entropy (RCMFDE), adopted as indicator framework for identification established in this paper. Firstly, decomposed by complete ensemble empirical mode decomposition with adaptive (CEEMDAN) method, intrinsic function (IMF) components are extracted. Secondly, RCMFDE IMF computed. Finally, hybrid optimization support vector machine (SVM) classify features, Relief-F filter utilized reduce dimension, particle swarm (PSO) wrapper optimize parameters SVM. The experimental results demonstrate encouraging accuracy apply approach states, an 91.11%.