A Generalized Data Driven Prognostic Algorithm Based On A Kalman Filter Mixture-Of-Experts Framework

This research is centered on the investigation of a modular methodology for application of adaptive Kalman filtering using a mixture-of-experts network to a data driven prognostic system for the Airborne Laser (ABL). In this framework, each expert nodewill be a Kalman filter modeled with failure modalities based on physical ABL system parameters. This framework provides the robust on-line prognostic capability for the complex systems of the ABL which will require an ability to estimate time-varyingsystem states with specified uncertainty and to rapidly react to changes in ABL system failure modalities and as well as occurrences of un-modeled failures. The use of data driven prognostic algorithms that can accurately predict time to faildistributions for key subsystems and functional elements of the Airborne Laser will allow timely remediation of failure modalities prior to their physical manifestations impacting overall system performance.