The development of an advanced diagnostic/prognostic system for the RB199 aero engine

Military aero gas turbine engines are highly complex engineering systems, which are extremely expensive to maintain and operate. In view of the increasing financial constraints facing fleet managers, there is a strong emphasis on reducing cost of operation without impacting performance or availability. In particular, in-service support experience with the RB199 engine has been highlighted within the UK MoD Propulsion Research Strategy (2003), and identified as the target for affordability improvements. For the RB199, as for most other aeroengines, health monitoring forms an essential part of the fleet management process. The existing health monitoring system for the RB199 includes the recording of a number of health indicators, which are used to inform decisions regarding engine rejection and overhaul. Under current fleet management practice, each of these different indicators is assessed against a fixed, predetermined threshold, the exceedance of which triggers engine rejection. The work reported here is part of an ongoing research programme to develop a more advanced diagnostic/prognostic capability for the RB199 fleet. The basis for the programme is the idea that, if the different health indicators could be combined in some suitable way, a much clearer view of the state of the engine could be obtained. Moreover, the use of Bayesian networks has been identified as providing an ideal mechanism through which such a synthesis can be obtained. To this end, the various available data sources relating to the engine condition have been identified, and a number of data handling and data processing issues have been addressed to allow the information to be produced in a appropriate form. On this basis, a Bayesian network representation of the engine has been developed, which allows all the observable data obtained from the engine to be incorporated simultaneously