Leveraging Gaussian mixture model for diesel engine fault diagnostics through oil filter debris analysis versus conventional oil analysis

Predictive maintenance strategies using oil analysis are highly effective for monitoring mechanical and industrial equipment health. However, oil filters can trap wear particles and make oil condition monitoring less reliable. To address this, analyzing particles in oil filters has emerged as a valuable complementary technique in this research. This study compares diesel engine oil analysis and oil filter analysis using the observational ferrography method (ASTM-D7919-14) on 50 mining equipment samples. Observational ferrography identifies particle type, size, and quantity in filters. Particle density is then classified using the Gaussian Mixture Model (GMM). Results reveal that in 20% of the examined cases, oil analysis shows acceptable results, while filter analysis detects severe engine wear, emphasizing the need for further inspection. This discrepancy underscores the importance of filter analysis in identifying hidden issues. The study recommends combining oil and filter analyses to enhance predictive maintenance accuracy and robustness. Additionally, the economic value of filter analysis is evaluated, offering a cost-benefit assessment of the integrated approach. By demonstrating the complementary nature of these techniques, the research promotes a more holistic predictive maintenance strategy, ultimately improving equipment reliability and operational efficiency.