Predictive maintenance practices represent a relevant feature for improving the useful life of the systems and decreasing costs. Hence, the modelling of the Health Indicator is a useful support in order to define the Remaining Useful Life of a system. In this work, an approach for the Health Indicator modeling of an oil refinery sub-plant is proposed; in addition, the Association Rule Mining is applied, in order to identify the components frequently requiring a work order prior to a stoppage of the plant: in this way, the Remaining Useful Life determined via the Health Indicator is used to inspect such components and, possibly, avoid the stoppage.