Intelligent Software Agents for Data Driven Prognostics

Impact Technologies, in collaboration with the Georgia Institute of Technology, proposes an innovative data driven prognostic system based on the integration of an Anomaly Reasoning Agent (ARA), Diagnostics Reasoning Agent (DRA), and Prognostic ReasoningAgent (PRA). With an Airborne Laser (ABL-based) focus, data acquired from a multi-sensor array is first validated, fused and key diagnostic features are extracted from it that are used in a high-level Support Vector Machine classifier. This conditionedinformation is then assessed by the ARA, which trigger events for the downstream diagnostic and prognostic agents. Once triggered, the DRA will utilize data driven models and failure mode propagation knowledge associated with the subsystem for determiningthe root cause of the anomaly. Also event driven, the PRA will predict the time to mechanical failure or time to conditional failure of a component or components within a subsystem given all data, features and data-driven models from the ARA and DRA.The PRA will implement integrated data-driven and physical (when available) models that exist at various levels of complexity. Within the system architecture described herein, the Prognostic Reasoning agent's function is to intelligently utilizeanomaly/diagnostic results, extracted features, experienced-based information and statistically estimated future conditions to determine the remaining useful life of the component or subsystem of the ABL system. The proposed data driven prognostic systemwill be demonstrated in an integrated software environment using simulated ABL data. The potential commercial use of the marketable data driven prognostic technologies is broad. The successful completion of the proposed work will lead to significantbenefits in a wide variety of areas, with a very substantial potential for commercial impact. This is primarily due to the fact that the Agent-based architecture is data-driven, hence allowing practical implementation for any type of complex system beingmonitored with low/high bandwidth data sources. Examples of key industrial customers that could benefit through use of the developed fault diagnostic technologies include commercial airlines, electric power producers, oil and gas transmission companies,and marine propulsion applications.