In-Flight Diagnosis and Anomaly Detection

In flight diagnosis and anomaly detection is a difficult challenge that requires sufficient observation and real-time processing of health information. Our approach uses formalized attributes that are available as selectable and enforceable properties necessary for diagnosis based on principles of model based engineering (MBE). Using this information, two strategies are proposed. The first is to use the concept of perfect detectors as executable assertions to verify at run-time correct operating envelope behavior. This information is used to check for correct behavior status or identify entry into a chain of events that could have failure impact. The proposed Phase I effort uses a combination of tool support to analyze the system, identify the properties to be checked, and the failure path information needed by the in-flight diagnosis service. This approach, is relevant to lowering the cost of systems since and provides important benefits related to V&V of complex systems.