Turbine Engine Performance Estimation Using Particle Filters

Development of a nonlinear particle filter for engine performance is proposed. The approach employs NASA high-fidelity C-MAPSS40K engine model as the central element, and addresses the issue of lack of observability of some of the engine health parameters in previous Kalman filter formulations. Proposed approach does not require linearity of the dynamics or Gaussian noise assumptions for satisfactory operation. The feasibility of real-time implementation of the proposed approach will be demonstrated using commercial, off-the-shelf General Purpose Graphical Processing Units. Phase I feasibility demonstration will show that the particle filter formulation of the engine performance monitoring system can overcome the limitations of previously employed approaches. Phase II research will develop a prototype implementation for hardware-in-loop simulations and eventual flight test.