<![CDATA[This study develops a machine learning-based predictive maintenance model for chemical tankers with capacities of 11,000–25,000 DWT using synthetic log-book data representing manual engine-room records without additional sensors. The XGBoost model predicts potential system failures within 14 days, achieving an Area Under Curve (AUC) of 0.9, recall of 0.82, and precision of 0.87. SHAP interpretability analysis identifies exhaust gas temperature differentials between cylinders, scavenge air pressure, and iron content in lubricating oil as the most influential predictors of failure. Implementation of the predictive system improves Mean Time Between Failure (MTBF) by 25.5% and system availability from 94.6% to 97.8%. Economic evaluation yields a Net Present Value (NPV) of USD 2.45 million per vessel with a Payback Period of 11 months. The findings confirm the reliability of machine learning-based predictive maintenance using operational data without expensive sensor infrastructure, supporting both efficiency gains and digital transformation within the maritime industry.]]>
Copyrights © 2025
