Condition monitoring of engine lubrication oil of military vehicles: a machine learning approach

Lubrication oil plays an important role in maintaining the health and performance of a land vehicle engine. Accurate condition monitoring of lubrication oil enables an effective predictive maintenance regime to be established. This can extend engine life as well as reduce over or under-servicing and other unnecessary maintenance costs. Machine learning models are useful for mining meaningful patterns from data samples. In this research, through the application of such models, we classify the condition of engine lubrication oil based on data from the Vehicle Health and Usage Monitoring System and laboratory test results of lubrication oil from a cohort of military land vehicles. The oil condition is classified into three categories: normal, degraded, and unsuitable. Feature selection methods are used to identify the best feature set for representing the lubrication oil condition. Importantly, the machine learning models employed provide the predicted output with justification in the form of explanatory rules pertaining to the lubrication oil condition. The findings indicate that (i) a good feature selection method is necessary to reduce the dimensionality of the feature set used for classification; (ii) machine learning provides a viable method for classifying oil condition with understandable justifications.