Defining a data-driven maintenance policy: an application to an oil refinery plant

Purpose – The purpose of this paper is developing a data-driven maintenance policy through the analysis of vast amount of data and its application to an oil refinery plant. The maintenance policy, analyzing data regarding sub-plant stoppages and components breakdowns within a defined time interval, supports the decision maker in determining whether it is better to perform predictive maintenance or corrective interventions on the basis of probability measurements.
Design/methodology/approach – The formalism applied to pursue this aim is association rules mining since it allows to discover the existence of relationships between sub-plant stoppages and components breakdowns.
Findings – The application of the maintenance policy to a three-year case highlighted that the extracted rules depend on both the kind of stoppage and the timeframe considered, hence different maintenance strategies are suggested.
Originality/value – This paper demonstrates that data mining (DM) tools, like association rules (AR), can provide a valuable support to maintenance processes. In particular, the described policy can be generalized and applied both to other refineries and to other continuous production systems.