Predicting tribological performance of liquid lubricants and coated-surfaces by machine learning approaches

Abstract

Minimizing friction and wear between contacting surfaces is a central focus in tribology, especially in improving lubricants' performance for engine durability and reliability. Low-viscosity lubricants, while beneficial for fuel efficiency, present challenges in reducing friction and wear. The shift to low-viscosity lubricants, such as SAE 5W30 and SAE 0W20, raises concerns about increased surface contact and potential durability challenges, highlighting the need for advanced additives like graphene and fullerene, as well as coatings, to mitigate these issues, while leveraging machine learning to predict tribological performance and reduce resource-intensive experimental trials. Graphene and fullerene have gained attention as potential additives to enhance the tribological properties of lubricants This study investigates the effects of 0.005 wt.% graphene and fullerene in SAE 0W20 lubricant, along with six different coatings (TiN, AlTiN, CrN, TiAlSN, TiCN, and DLC-AlTiN) on steel block. The primary aim is to develop a dataset that explores the relationships between these additives and coatings with critical tribological parameters—coefficient of friction, wear area, and temperature. Using a rod and ring tribometer, this study measures the tribological performance, with SEM and EDX analyses for surface characterization. The findings indicate that graphene and fullerene additives significantly reduce friction and wear, with graphene showing better heat dissipation. The coatings, particularly CrN, reduce wear and friction, although CrN showed higher wear compared to others. Machine learning models, including Support Vector Regression (SVR), Random Forest (RF), and Decision Tree (DT), were used to predict the tribological parameters. RF outperforms SVR and DT in predicting both friction and wear area, while DT excels in predicting temperature. This research contributes to improving lubrication strategies and predictive modeling, offering valuable insights for enhancing engine performance and durability in automotive engineering.