As the demand for ocean exploration increases, studies are being actively conducted on autonomous underwater vehicles (AUVs) that can efficiently perform various missions. To successfully perform long-term, wide-ranging missions, it is necessary to apply fault diagnosis technology to AUVs. In this study, a system that can monitor the health of in situ AUV thrusters using a convolutional neural network (CNN) was developed. As input data, an acoustic signal that comprehensively contains the mechanical and hydrodynamic information of the AUV thruster was adopted. The acoustic signal was pre-processed into two-dimensional data through continuous wavelet transform. The neural network was trained with three different pre-processing methods and the accuracy was compared. The decibel scale was more effective than the linear scale, and the normalized decibel scale was more effective than the decibel scale. Through tests on off-training conditions that deviate from the neural network learning condition, the developed system properly recognized the distribution characteristics of noise sources even when the operating speed and the thruster rotation speed changed, and correctly diagnosed the state of the thruster. These results showed that the acoustic signal-based CNN can be effectively used for monitoring the health of the AUV’s thrusters.

Static eccentricity is an important early-stage fault in marine diesel generators, as small air-gap deviations caused by misalignment or mechanical wear can escalate into bearing damage and rotor–stator contact. To address the challenge of detecting such subtle faults, this study proposes a current signal analysis method based on the two-dimensional short-time Fourier transform (2D STFT) for early detection of static eccentricity faults in marine diesel generators. Using three-phase currents measured during normal operation and fault data synthesized with a physics-based electromechanical coupling model (1–5% eccentricity), we construct a two-dimensional phase–time representation rather than treating each phase as an independent one-dimensional time series and then apply 2D STFT. This formulation enables the simultaneous capture of inter-phase relationships and spatial patterns in the time–frequency–phase domain. Experiments indicate a distinct energy rise near 1020 Hz as static eccentricity increases. This trend enables the proposed method to distinguish small faults of approximately 5% eccentricity, which remain difficult to detect using conventional 1D STFT. As a result, the approach improves the diagnostic accuracy of non-contact, current-based monitoring for static eccentricity faults. Future work will include validation using real in-service fault data and extensions to other fault modes such as dynamic eccentricity and bearing defects.