Reliability Assessment of Predictive Maintenance Techniques for Marine Propulsion Systems

The maintainability and reliability of marine propulsion systems is critical for the operational efficiency, safety, and economic feasibility of maritime operations and activities. This study aims to access the reliability of predictive maintenance techniques employed for early detection of defects in marine propulsion systems to reduce breakdowns and downtime. The model integrated a reliability framework with field data gathered from the case study of Tugboat Abun to facilitate risk-informed repair prioritization for the reliability of the vessel's marine propulsion system. This will use the ANN‑based prediction model with FTA and FMECA for risk‑informed maintenance prioritization towards reliability of the marine propulsion system and assess the effectiveness of the combined ANN‑FTA‑FMECA framework in reducing downtime, extending equipment service life, and lowering operational costs on marine propulsion systems The model's prediction for the injectors of the starboard propulsion plant of the vessel averaged 95 percent, with a mean square error of 12.13 and a mean absolute error of 3.13, signifying that the predictions were, on average, highly accurate relative to the actual values. The subsystem failure probability graph shows the relative vulnerability of each component group contributing to the overall propulsion system unreliability. The highest probability was observed in the fuel system (P = 0.1817), indicating it as the most failure-prone subsystem due to frequent occurrences such as injector and filter blockages. This was followed by the engine block assembly (P = 0.1299) and main engine cooling system (P = 0.0672), both critical for maintaining propulsion stability and thermal regulation. Moderate failure probabilities were recorded in gearbox cooling (P = 0.0580) and main engine lubricating oil system (P = 0.0390), signifying poor cooling and inadequate lubrication (viscosity loss). Subsystems such as shafting (P = 0.0065), coupling, bearing, and propeller (P = 0.0000) show near-negligible probabilities, implying effective structural reliability. Overall, the combined subsystem probabilities yielded a top event probability (Ptop = 0.5882), suggesting above-average overall system reliability. The results demonstrate that the Artificial Neural Network predictive model, utilizing the multi-layer perceptron architecture, can accurately forecast marine machinery breakdown, hence improving dependability and reducing downtime.