Engine Diagnostic, Trend Monitoring, and Life Management

The purpose of this program is to demonstrate a novel approach to condition-based diagnosis and life management for gas-turbine engines. The long term goal is to eliminate heavy, labor intensive on-board computer hardware by using "educated", aircraft-specific, neural network computer "chips" of diagnostic software, capable of intelligent interpretation of both engine sensor signals and aerothermal performance details. The proposed approach uses recent advances in neural network technology, finite element based rotor dynamic analysis, and throughflow performance analysis. Phase I will involve the development of neural software for structural diagnosis of gas turbine engines and the demonstrtion of this software for engine diagnostic purposes using a physical rotating model. This work will demonstrate how dedicated finite element modeling and select engine measurements may serve as the "brains" for a neural network based diagnostic system. Engine performance aspects will be accomplished with the same neural network approach by replacing the finite element model with an engine performance model. During Phase II, prototype aircraft monitor will be developed in conjunction with Rolls-Royce Incorporated by extending the principles demonstrated in Phase I. The neural network based system will be "trained" in the dynamic and mechanical behavior of an engine, and in the engine performance characteristics as determined by aerothermal cycle synthesis. The prototype system will provide the USAF with demonstrated, real-time estimates of remaining component life and of engine performance for a select engine, based on real-time, updated behavior during operation.