Akademik Veri Yönetim Sistemi
SELCUK 13TH INTERNATIONAL CONFERENCE ON APPLIED SCIENCES, Konya, Türkiye, 12 - 14 Aralık 2025, ss.294-306, (Tam Metin Bildiri)
Marine engines are critical components of maritime transportation systems, and their reliable operation is essential for safety and efficiency. This study presents a comparative analysis of five ensemble learning methods for fault diagnosis in marine engines: Boosted Trees (AdaBoost), Bagged Trees, Subspace Discriminant, Subspace KNN (K-Nearest Neighbors), and RUSBoosted Trees. A dataset containing 10000 samples with 19 sensor features is used, covering normal operation and seven fault classes, including fuel injection fault, cylinder pressure loss, exhaust gas overheating, bearing/vibration fault, lubrication oil degradation, turbocharger failure, and mixed faults. The performance of each method is evaluated using the accuracy metric, and the validation and test results are additionally analyzed in terms of confusion matrices. Results show that the Bagged Trees ensemble model achieved the highest test accuracy of 87.8%, indicating that ensemble methods are effective tools for predictive maintenance and condition monitoring of marine engines.