A Fault Diagnosis System for Turbo-Generator Set by Data Mining

Aiming at difficulties of vibration fault diagnosis for turbo-generator sets, an intelligent data-mining system based on acquired data in SCADA systems is structured. The hard core of the system is a focusing quantization algorithm and a reduction algorithm. The focusing quantization algorithm put focus on the transition point from normal to abnormal state of variables, the resolution near the focus is enhanced to improve diagnostic accuracy. The reduction algorithm based on rough set theory is used to find the minimal set in all preparation variables. The diagnosis rules mining from SCADA systems´ database are expressed directly by variables in database, so it is easy to understand. A vibration fault diagnosis system for 600MW turbo-generator set is designed by the proposed approach, its running results in a thermal power plant of Guangdong Province showed that the system had high accuracy and satisfied fault diagnosis requirement of large-scale turbo-generator set