On-Board Real-Time Generator Component Failure Diagnostics

The advent of prognostic health management (PHM) philosophies in condition-based maintenance (CBM) has clear benefits in cost savings and improved readiness. Generator failures in the E2C fleet have been a continuing problem, and this component wouldbenefit from a CBM-PHM approach to maintenance. Since real-time data is needed for PHM models, fiber optic sensors could be used to eliminate the noise associated with electrical sensors inside an electrical generator. Prime Research is working with amajor US research university to develop a novel PHM system which incorporates both fiber optic sensors and neural networks to prognosticate reliability and potential failures. The novel predictive system incorporates constantly updated data from design,manufacturing, quality control, and recent operational events into a model with a neural network component that can be `trained' to adjust to current events in any particular generator, and feed that data back into a fleetwide database to immediatelyupdate the prediction system of the entire inventory. The proposed Phase I program will demonstrate the feasibility of incorporating fiber optic sensor data into a PHM system for E2C generators, drastically improving remaining life predictions.