Application Of Machine Learning Methods in Detecting Anomalies On L.O Cooler Component through Hyperparameter Optimization

Abstract

LO Cooler is a crucial component in the ship's engine that functions to maintain the temperature of the lubricating oil within safe operational limits. Interference or failure of the LO Cooler can cause serious damage to the vessel's main engine. Early detection of anomalies in LO coolers can help prevent greater damage, thereby improving the reliability and operational efficiency of the vessel. This study proposes a machine learning-based anomaly detection method to monitor the performance of the LO Cooler in real time. In this study, operational data from the RPM, temperature, and pressure, on the LO Cooler were collected over a period of time. Machine learning algorithms, such as Support Vector Machine (SVM) and Decision Tree are applied to detect anomalous patterns in the data that indicate potential failures. The result that Decision tree modeling is the most accurate method for developing models for the LO Cooler of the main engine. With a data split ratio of 90:10 for training and testing and it achieved the highest accuracy, as indicated by the best following metrics MAE, RMSE, RAE, TP Rate and F-measure.