Advanced modeling of propeller tip-vortex cavitation noise using deep learning

Propeller tip-vortex cavitation (TVC) noise significantly influences naval vessel acoustic signatures. However, existing cavitation noise models in underwater acoustic simulators fail to account for characteristics of propeller TVC noise. This study proposes an advanced method that integrates a probabilistic representation of noise occurrence with generative adversarial networks (GANs) to generate realistic propeller TVC noise. Experimental data collected under controlled cavitation tunnel conditions were used to model the stochastic occurrence of TVC events based on propeller revolutions per second (rps), while GANs generated high-frequency, impulsive waveforms. The proposed method was validated through comprehensive evaluations, including statistical analyses of waveform similarity (cross-correlation and low-dimensional embedding), time-frequency analyses using spectrograms, and auditory feature analyses based on cepstral coefficients to assess perceptual fidelity. The results demonstrated that the proposed approach effectively captures both the stochastic and acoustic properties of TVC noise, thereby improving the realism of underwater acoustic simulations and outperforming existing methods. [This work was supported by the Korea Research Institute for Defense Technology Planning and Advancement (KRIT) grant funded by the Korea government [Defense Acquisition Program Administration (DAPA)] (No. KRIT-CT-22-052, Physics-guided Intelligent Sonar Signal Detection Research Laboratory, 2025)]