Predicting engine oil degradation across diverse vehicles and identifying key factors

Engine oil degradation accelerates wear and corrosion in engines, leading to failures and reduced performance. Traditional maintenance methods are inefficient, either replacing oil too frequently or risking engine damage due to delayed changes. This study proposes a novel machine learning–based predictive maintenance system that accurately forecasts engine oil degradation across various vehicles using readily available vehicle information and driving data, such as engine mileage and displacement. Over four years, 820 oil sampling data were collected from 169 commercial buses and trucks to analyze engine oil properties and operational parameters. Machine learning regression models including Support Vector Machine (SVM), Random Forest (RF), and Gaussian Process Regression (GPR) were developed to predict key oil properties such as base number, conradson carbon residue, iron content, and viscosity. The GPR model demonstrated superior predictive accuracy, effectively capturing complex relationships in the data. Shapley value analysis identified engine oil usage mileage, engine displacement, total engine mileage, and monthly mileage as significant factors affecting oil degradation, while vehicle size and bus/truck type were found to have low importance. The proposed system enhances predictive maintenance by accurately predicting oil health, improving maintenance efficiency and reducing costs. This approach offers a robust solution for monitoring engine oil condition across various vehicles, ensuring optimal engine performance and longevity without the drawbacks of traditional methods.