A railway is an extremely complex system requiring maintenancedecision support systems to gather data from many disparatesources. These sources include traditional maintenanceinformation like condition monitoring or work records, as well astraffic information, given the criticality of maintenance inavoiding traffic disruptions and the need to minimise the trackpossession time for maintenance.A methodology is required if maintainers are to understand thedata as a whole. Context engines try to link the various dataconstellations and to define interactions within the railwaysystem. This is not easy since data have different natures, originsand granularity. But if all information surrounding the railwayasset can be considered, decisions will be more accurate andproblems like false alarms or outlying anomalies will be detected.The contextualisation of the data seems to be a feasible way toallow condition monitoring data i.e physical measurements andother variables, to be understood under certain conditions(weather, regulations etc.) and as a consequence of certain actions(maintenance interventions, overhauls, outsourcing warrantiesetc.).This paper proposes the use of context engines to providemeaningful information out of the overwhelming amount ofcollected and recorded data so that proper maintenance decisionscan be made. In this scenario, fluffy information coming fromwork orders and expertise of maintainers is a big issue since suchinformation must be converted to numerical values. The fuzzylogic approach seems a promising way to integrate suchinformation sources for diagnosis.