Prediction of Lubrication Oil Parameter Degradation to Extend the Oil Change Interval Based on Gaussian Process Regression (GPR)

Abstract

In this work, the degradation of selected lubrication oil parameters until the specified threshold is predicted based on Gaussian process regression (GPR) to extend the oil change interval. Kinematic viscosity (40°C) and total acid number (TAN) was selected based on Mahalanobis-Taguchi Gram-Schmidt (MTGS) analysis. Kinematic viscosity (40°C) degradation prediction model presented a better result with an R squared of 0.61835, RMSE of 0.97142 and predicted maximum oil age of 31.9 months. The predicted maximum oil age value is validated by modelling the degradation of remaining lubrication oil parameters namely water content and particle counts, these parameters are within its threshold value at 31.9 months. The study proves that the degradation of kinematic viscosity (40°C) can be predicted based on GPR modelling. The corresponding oil age at threshold of parameter was identified to be longer than the actual duration of lubrication oil in service and the oil change interval can be extended beyond 26 months.

Keywords

• lubricant condition monitoring
• predictive maintenance
• statistical method
• mahalanobis-taguchi gram-schmidt
• gaussian process regression