SBIR-STTR Award

This effort will result in the development of a Satellite Telemetry Anomaly Resolution (STAR) tool that automatically: 1) analyzes historical telemetry data, 2) learns expected signal characteristics for specific telemetry signals using prediction networksä (an advanced statistical modeling approach) and innovative case based reasoning (CBR) techniques, and 3) applies these models to detect and isolate anomalies. It will reduce mission operations data analysis costs, enhanced the ability of operators to handle the large volume of incoming data, and reduce the likelihood of soft failures being undetected until they cause significant damage to valuable spacecraft components. It will also be much more cost-effective than expert system or model based systems. This effort will demonstrate the feasibility of developing vastly superior satellite diagnostic and prognostic systems that: 1. Increase fault detection and isolation well beyond that provided by conventional built-in test (BIT), rule-based expert systems, and model-based systems. 2. Distinguish and isolate sensor failures from equipment failures. 3. Estimate system degradation. 4. Predict an important class of impending failures sufficiently in advance to avoid them. A major goal is to enable STAR to be implemented so inexpensively for the value offered that it will quickly become popular on many different satellites.|

Benefits:
This technology can be used in monitoring commercial satellites, instruments in hospital intensive care units, machines in factories, and numerous other aerospace, healthcare, and manufacturing applications.|

Keywords:
Modelquest Expert Case Based Reasoning (Cbr) Case Libraries Diagnostics Satellite Telemetry Signal

The proposed Phase II project will result in the development of a Satellite Telemetry Anomaly Resolution (STAR) product that automatically: 1) analyzes historical or simulated telemetry date, 2) models expceted signal characteristics using an innovative mix of statistical network [TM] (AbTech's advanced statistical modeling approach), case based reasoning (CBR), pattern recognition, and machine learning techniques, and 3) applies these models to detect and isolate anticipated and unanticipated anomalies. It will reduce mission operations data analysis costs, enhance the ability of operators to handle the large volume of incoming data, and reduce the likelihood of soft failures being undetected until they cause significant damage to valuable spacecraft components.During Phase I, AbTech developed an innovative satellite anomaly detection prototype which uses a hybrid approach of statistical network technology and CBR. Preliminary results suggest that very high anomaly detection capabilities (>98%) can be achieved in a very automated and inexpensive fashion. Initial studies indicate that the statistical networks used in STAR willbe able to detect a large number of diverse and subtle anomalies.By integrating statistical network, CBR, and other modeling technologies into an integrated tool designed for diagnostics, AbTech can create the most advanced and automated satellite telemetry monitoring capability available.