This report describes the status of an on-going effort to develop software capable of detecting sensor failures on rocket engines in real time. This software could be used in a rocket engine controller to prevent the erroneous shutdown of an engine due to sensor failures which would otherwise be interpreted as engine failures by the control software. The approach taken combines analytical redundancy with Bayesian belief networks to provide a solution which has well defined real-time characteristics and well-defined error rates. Analytical redundancy is a technique in which a sensor's value is predicted by using values from other sensors and known or empirically derived mathematical relations. A set of sensors and a set of relations among them form a network of cross-checks which can be used to periodically validate all of the sensors in the network. Bayesian belief networks provide a method of determining if each of the sensors in the network is valid, given the results of the cross-checks. This approach has been successfully demonstrated on the Technology Test Bed Engine at the NASA Marshall Space Flight Center. Current efforts are focused on extending the system to provide a validation capability for 100 sensors on the Space Shuttle Main Engine.
Real-time sensor data validation
1994-04-01
Report
No indication
English
Real-Time Sensor Data Validation
NTIS | 1994
|Real-time sensor validation for propulsion systems
AIAA | 1998
|Real-Time Sensor Validation for Propulsion Systems
British Library Conference Proceedings | 1998
|Real-Time Sensor Validation for Autonomous Flight Control
British Library Conference Proceedings | 1997
|