Estimating failure times of ship main engine combustion parts

Incorrect maintenance practices in marine main engines, along with lubrication oil and fuel-related issues, are the primary factors leading to major failures. These failures not only result in significant financial losses but also negatively impact navigational safety, especially considering the maritime traffic conditions at the time of failure (e.g., during maneuvering, canal or strait transits). This study aims to statistically identify the main engine and fuelrelated factors contributing to engine failures and to forecast the timing of such failures. According to the statistical findings, the factors causing these failures include the sulfur content in the fuel, vanadium, catalytic fine particles, and micro carbon residue. Due to the failure times in the dataset not exhibiting a normal distribution, the desired correlation analysis value could not be achieved in temporal prediction using Artificial Neural Networks (R = 0.67289). Therefore, other machine learning algorithms were utilized via the WEKA software for condition-based fault prediction (i.e., presence or absence of failure). As a result, the Random Forest model achieved a high overall accuracy of 94.7% in conditionbased fault prediction.