Identification of faults in turbines by analysis of pressure signals, body noise and shaft vibrations

Among the critical components in power stations are the turbines, as they have moving parts, which are highly stressed. Because of the high costs which can be caused by shutdown of a turbine, one tries to recognise faults early. The author is concerned here with methods of investigation, which are particularly suitable for monitoring running blades and fixed blades of turbines. The main point of the investigations concerns turbines on which various faults have been simulated. The types of detectors used are also tested, in order to see which types are most suitable for monitoring turbines. The signals of the detectors are evaluated by different models and mathematical analyses. Apart from formulating the criteria for recognising damage, the limits of sensitivity for the methods developed are determined. Finally, a program system is introduced for the methods produced and an automatic monitoring process is therefore created. Using this automatic monitoring system, characteristic features of the signals of different detectors are processed and the state of the machine is decided by a 'decision tree', using these characteristics. (orig.)