ABSTRACT
III. Components that can be fixed (with different repair procedures according to the entity of the damage) are sent back to the costumer in an As Good As New (AGAN) state.
With regards to step II, currently GE carries out the reliability life data analysis by modeling the component behavior as binary, i.e., there are only two possible component states: ‘working’ and ‘failed’. This is done by treating ‘Good’ as ‘working’ state and combining ‘Light’’, ‘Medium’, ‘Heavy’ into a single ‘failed’ state (Fig. 2). In this work, we also adopt this latter simplified binary
1 INTRODUCTION
General Electric (GE) offers to its customers a maintenance service on a set of components, such as the nozzles of the turbines, which is based on the subsequent steps:
I. The customer sends GE the component removed from the system upon a maintenance inspection. GE experts check the component and qualitatively classify its health state into ‘Good’, ‘Light’, ‘Medium’ and ‘Heavy’ (see Fig. 1).
