Abstract: | TIGER, the knowledge based system developed by Intelligent Applications Ltd, is a real-time system that monitors a gas turbine at once per second intervals. It includes a number of techniques for fault detection and fault diagnosis. However, techniques for event detection have not been developed as much as other techniques applied by TIGER. This dissertation presents an event detection system which successfully isolates interesting events from noise in a constantly varying environment. To perform such task, the event detector (ED) system applies some statistical tests on continuous data. A mixture of signal processing techniques, combined with statistical features and a rule-based approach, enables the ED system to detect, classify and associate significant events.
|