Data Driven Prognostics

We propose to develop a prototype suite of recursive partitioning (RP) methodologies to provide advanced warning of failure, malfunction and/or performance degradation for Airborne Laser Program subsystems in service. Through better monitoring and analysisof the status of its systems, prediction of equipment life remaining can be made and corrective action taken to enhance the availability of the Airborne Laser System and increase its cost-effectiveness. In general, the cost of failure of military equipmentcan be enormous in terms of resources, compromised mission success, and potential loss of life. Providing early warning before equipment failures transpire has a commensurately huge economic benefit. Our proprietary RP algorithms are already commercializedin the pharmaceutical industry in software used to predict patient outcomes in clinical trials. Modifying our methods to a hazard-based RP framework will enable us to meet MDA prognostic needs. The risk associated with this work is nominal but it ispossible that the premise we are working from will prove inadequate and require further investigation. With the funds from this grant we propose to further develop our patented recursive partitioning algorithms to perform hazard RP in a novel applicationto provide accurate predictions of system function/malfunction. The cost of failure of military equipment can be enormous, both in terms of resources and possible loss of life. Providing early warning before failures happen can reduce these costs. TheMDA anticipates the use of such successful technology with the Airborne Laser Program. Beyond applications to equipment failure, the technologies have applicability to human subjects in predicting lifetimes of people under various conditions.