Recent advancements in predictive maintenance for ship propulsion engines could revolutionize the maritime industry, significantly reducing operational costs and enhancing reliability. A study led by Jinkyu Park from the Division of Marine System Engineering at Mokpo National Maritime University in South Korea has developed an innovative machine learning-based algorithm designed to detect anomalistic symptoms in ship engines before they lead to failures.
As ships operate for extended periods—often 20 to 30 years—the costs associated with maintenance can be substantial, sometimes accounting for 10 to 30% of total management expenses. The financial impact of improper maintenance and unexpected breakdowns can further exacerbate these costs. Park’s research addresses these challenges by focusing on predictive maintenance (PdM), a strategy that monitors the condition of machinery to preemptively plan maintenance activities.
The algorithm developed in this study utilizes data collected from the alarm monitoring system of a ship, analyzing various operational parameters such as revolutions per minute (RPM), engine load, and exhaust gas temperature. By identifying key indicators, including cylinder exhaust gas temperature (CET) and maximum cylinder combustion pressure (CCP), the algorithm can effectively predict potential engine anomalies.
Park stated, “To intuitively detect the anomalistic symptoms of the propulsion engine, the main engine condition criterion value (MCCV) was defined based on the CET and maximum CCP data.” However, the initial MCCV approach faced challenges due to the variability in operating conditions. To overcome this, a revised MCCV (RMCCV) was introduced, which better reflects the changing conditions of the engine and the marine environment.
The results of the study are promising, with the RMCCV showing high explanatory power for predicting engine conditions, with adjusted R-squared values between 0.754 and 0.968. This indicates a strong correlation between the algorithm’s predictions and actual engine performance. The mean absolute error (MAE) for the CET and maximum CCP also demonstrated excellent predictive performance, ranging from 4.095 to 7.411 and 1.188 to 1.867, respectively.
The commercial implications of this research are significant. Shipping companies can reduce maintenance costs and downtime, leading to more efficient operations and increased profitability. By implementing such a predictive maintenance system, companies can shift from reactive maintenance strategies to proactive ones, ensuring that engines are serviced before issues escalate into costly repairs.
Moreover, the study highlights the potential for developing a comprehensive PdM platform capable of real-time anomaly detection and maintenance identification. This could lead to a new standard in maritime operations, where technology plays a crucial role in enhancing the safety and efficiency of ship operations.
As Park noted, the future plans include validating the algorithm on operational ships and creating a platform that integrates these capabilities. The findings, published in “Applied Sciences,” pave the way for smarter, data-driven approaches to maritime maintenance, ultimately benefiting the industry and its stakeholders.