Abstract:

FMEA is a useful tool that helps to find possible flaws in a system or element, mainly in the design phase. But it is a mere document that has no further 'direct' use than the observation. Mapping all this knowledge into a Bayesian Network would make it possible to use the information in further ways, like software applications for maintenance. A good FMEA has the necessary features to build a good Bayesian Network: Bottom-up (or Top-Down) analysis of all the components and subcomponents and cause-failure-effect chains. In this paper we will detail the steps followed to create the Bayesian Network, using the FMEA of a Marine Diesel Engine, as well as the use of it in a diagnosis application that uses a set of on-line lube-oil sensors.