Predictive Maintenance of Hybrid Propulsion System

Abstract

Hybrid propulsion systems, integrating internal combustion engines with electric motors, represent a significant advancement in maritime technology, offering improved efficiency and reduced emissions. However, their complexity introduces challenges in maintenance and reliability. Traditional maintenance strategies are often inadequate for these dynamic systems, leading to unplanned downtime and increased costs. This research develops and validates a predictive maintenance framework specifically designed for hybrid propulsion systems in maritime applications, integrating vibration, thermal, and electrical data to enhance system reliability and reduce maintenance costs. The study employs advanced signal processing techniques including Root Mean Square (RMS), Kurtosis, Fourier’s Law, and Wavelet Transforms to extract degradation features from sensor data. Multi-sensor fusion is achieved using Dempster-Shafer evidence theory and weighted entropy-based models to resolve data conflicts and provide a holistic health assessment. Failure prediction and Remaining Useful Life (RUL) estimation are conducted using Proportional Hazards Models (PHM) and Weibull distributions. The framework was validated through case studies on two hybrid-powered vessels: a 2 MW coastal cargo ship (Ship A) and a 5 MW offshore support vessel (Ship B). Results showed that Ship A achieved an MTBF of 1,440 hours and 99.45% availability, while Ship B, operating under harsher conditions, recorded an MTBF of 864 hours and 99.08% availability. The PHM-based RUL estimation achieved a Mean Absolute Error of 12.5 hours (15.6% error), demonstrating high predictive accuracy. Economic analysis indicated a potential 40% reduction in annual maintenance costs compared to traditional methods.