This paper presents the development of a new fault diagnosis and prognosis algorithm based on qualitative modeling, which provides improved fault isolation. A fault diagnosis and prognosis algorithm is developed to detect and identify faults in the startup components of turbine engines while the faults are still in progress. Such diagnosis and prognosis will make it possible to take proper action before the system breaks down. The evidence associated with startup component failure is based on the aggregation of three dynamic events occurring in different time windows. These events are observable from speed at peak EGT (exhaust gas temperature), peak EGT, and start time. Discrete event modeling observes the unsynchronized occurrence of events. The algorithm was tested with data collected from the field; test results were obtained for twenty-nine engines, including six engines with failed startup components. The developed fault prognosis system successfully predicts failure for all six cases. In the earliest case, alarms triggered sixteen flights before the startup component breakdown and five flights in advance for the latest case.

This content is only available via PDF.
You do not currently have access to this content.