AUTOMATED LEARNING FOR REAL-TIME EXPERT SYSTEM IN MONTORING AND CONTROL

NASA/KSC is a center that is rich in needs for affordable, reusable software control systems. Population of knowledge for a control system has been labor-intensive and costly task prone to errors and omissions. This STTR addresses the need to automatically populate control system tools from historical archives. The use of Adaptive Machine Learning (AML) techniques has proven that we can populate the SCL Rule-Based Expert System with monitor and control rules for a Space Shuttle Main Engine data stream. This approach keeps the human in the loop but removes mundane tasks, and allows analysis in real-time, post-test, and post-flight.Our Phase I work focused on determining nominal and off-nominal behavior during real-time monitoring. In Phase II we will expand and refine this capability and add an advisory system coupled with a diagnostics engine. This will greatly enhance the system as an operator aid and provide insight into the source of the anomaly and suggest corrective actions. An expanded GUI capability for data signature analysis and real-time monitor and control is also planned. The system is envisioned as being applied to testsets and range systems to help automate equipment monitoring, aid in preventative maintenance and increase range capacity.