Advancing condition-based maintenance of naval propulsion systems with ensemble learning techniques

Abstract

In the field of industrial engineering, especially in the operation of Gas Turbine (GT) propulsion systems used in frigates, ensuring reliable and efficient performance is crucial. These propulsion systems enable ships to navigate quickly, respond swiftly to threats, and successfully carry out important missions. However, the harsh conditions at sea, such as saltwater corrosion and temperature extremes, cause significant wear and tear on these systems, which can threaten both their performance and the safety of the crew. This makes accurate fault detection in GT compressors and turbines essential. This research introduces a novel approach that combines machine learning (ML) with advanced ensemble learning techniques specifically tailored for diagnosing faults in GT propulsion systems. The key innovation lies in developing a hybrid framework that leverages both the strength of traditional ML classifiers and the robustness of ensemble methods, improving the accuracy of fault diagnosis while reducing the risks of overfitting. Unlike conventional methods that focus on one type of algorithm, our method integrates diverse models to achieve better generalization and stability under varying operational conditions. Through this analysis, we assess the performance and stability of different strategies, focusing on how well this hybrid ensemble approach works compared to standard ML approaches across a variety of classifiers. The research offers valuable insights and provides a solid framework for understanding the flexibility and effectiveness of this novel technique. By connecting theory with real-world applications, this study aims to significantly improve fault detection in vital naval propulsion systems, ultimately contributing to better performance and reliability in maritime defense operations.