Projects

Air Force SBIR on Engine Health Management

MitekAn has completed an 2008 Air Force Phase I SBIR project in development of information decision software tools for turbine propulsion prognostics and health management (PHM) systems. Air Force has recently selected proposal for Phase II SBIR continuation of the project as one it expects to award.

The project is managed by AFRL and is carried in collaboration with GE Aviation (GE Engines).  Phase I SBIR project developed algorithms for engine Prognostics and Health Management (PHM). The algorithms implement optimal estimation of the engine health state from a mixture of discrete and parametric data. The discrete data are BIT/BITE (built-in-test/built-in-test-equipment) flags set by low-level detection logic in engine controller. The parametric data come from engine gas path sensors. We call the proposed approach of optimal fault state estimation from mixed discrete and parametric data the Mixed Data Fusion. The fault states of the engine considered in this project include gas path fault events, such as bird strike or blade loss, and their intensities. We also consider LRU (Line Replaceable Unit) fault states that have gas path impact, such as offset errors in gas path sensors or actuators.  Finally, we consider LRU-related faults that do not have gas path impact but cause generation of the BIT/BITE codes. The proposed approach to Mixed Data Fusion is proven to be feasible. It can substantially improve the accuracy of diagnostics of root cause faults. The approach enables reduction of false alarms and fault ambiguities. The computations take a few milliseconds. The feasibility demonstration of the approach was done using a high-fidelity simulation model for F110-GE-129 engine. The proposed algorithms were demonstrated to reduce the false error rate substantially despite large noise in the parametric data and distortion of the discrete data. At the same time the algorithms are demonstrated to discriminate between many more different fault conditions as the existing discrete logic.